HH52-01-10-2021

Harpreet: [00:00:08] What's up, everybody? What's going on? Welcome to the artist Data Science. Happy hour. It is Data science. Happy hour number fifty two. That means 52 weeks of being here every single Friday at 4:30 p.m. Central time to just have a gathering of amazing Data heads. Then just talk. All sorts of amazing stuff, man. Thank you guys so much. I'm eternally grateful for every single one of you who show up and make these Data times happy hours possible. I literally it would not be what it is without you guys. Shout out to all the people who take time out of their busy schedules to come here and answer questions. Tom AIs Bean was just kanji. Thank you guys. Come in and help in and answering people's questions.

Speaker2: [00:00:57] You know, Dave

Harpreet: [00:00:57] Linger all the people that are here since pretty much the beginning. I'm eternally grateful to all of you guys. Data times happy. I would not be what it is without. You guys come in here and just drop in such amazing knowledge on everyone and to everybody else, everybody who comes here to learn and to grow and get better. Thank you guys for taking time out of your schedule to do make a commitment to improving. Man, I really, really appreciate you guys being here and making this happen, man like that. It's crazy. I did not did not think it would last this long, but it has every single Friday. And I think, you know, I think we're we're we're gearing up for another fifty two of these weeks, for sure. A year, and I know a lot of you have come a long way, I know a lot of folks here have. That came so far, man, like I mean, a lot of people learning new jobs, a lot of people, you know, get new opportunities and whatnot. And everybody's growing a little bit. I'm just wondering. How far have you come in the last 52 weeks? The last 52 weeks? Do you reflect back on where [00:02:00] you were last October at this time? How far have you come? How far have you come and where are you going in the next 52, two weeks? I want to know this is something that I I want to know. Man, what's going on? Where, how far have you come the last 52 weeks?

Speaker2: [00:02:13] Where are you going

Harpreet: [00:02:14] In the next 52? Let's start with Monica, and then we'll go to Eric. And by the way, as usual, you guys have questions. Let me know.

Speaker3: [00:02:25] Awesome. Thank you so much for starting with me, I actually can't stay, but I wanted to come in and say a huge, huge thank you putting this event on.

Speaker2: [00:02:35] It has been

Speaker3: [00:02:36] A pleasure to be here to meet such amazing people start to finish. I mean, I've just been on roller coaster with everybody else throughout these past two years, and it's been really, really wonderful to get together, you know, at least on a weekly basis with these wonderful people. So again, thank you. Thank you so much and congratulations. Fifty two weeks.

Harpreet: [00:02:58] Thank you. Thank you. Yes, congratulations to all of us here for making this happen and just creating this awesome space for folks. So before you go, Monica, let us know, man. Where do you got planned for the next fifty two weeks? What's going on?

Speaker2: [00:03:10] Where are you headed?

Speaker3: [00:03:11] Oh, the next fifty two weeks just to remain awesome. Continuous learning, of course. Learn something new every day to keep your mind fresh.

Harpreet: [00:03:22] Well, they absolutely love it. Let's go to Eric in the after Eric, whoever else wants to go, let me know. Monica, thank you so much for stopping by. Maybe after Eric can go to Kenji and then after Kenji, maybe we can hear from people who as somebody I've probably never heard from before. So I'd love to hear from you.

Speaker2: [00:03:37] Ashton, I

Harpreet: [00:03:37] Know Ashton's got some big news to share with us as well. Eric, how far you come the last two weeks?

Speaker4: [00:03:44] Where you headed. Okay. Yeah. Let's see, fifty two weeks ago. Beginning of October. List starting module two of the fall semester of the Masters, and so I'm just starting [00:04:00] into the. Decision trees and basketball, but we're not totally brand new. I did not know one thing. Ah, I still work with it and definitely and I was I was active on LinkedIn, but I definitely like

Speaker2: [00:04:23] More active,

Speaker4: [00:04:24] More tracks. So it's nice because I might be one of those people who can look back in year and say yes, but with that goes from being. Look in 52 weeks forward to be doing a lot of the same stuff I'm doing right now is because I have more experience with some of the especially like company specific stuff because I'm. And then the other piece is your year ago I've been pretty solid on kind of my personal personal brand is as a human. And a lot of ways. But I've been

Speaker3: [00:05:03] Really. That looks like some

Speaker4: [00:05:08] Of the next era.

Harpreet: [00:05:12] But. Right, right at the good part, Eric's microphone cuts out completely, your microphone has been shaky this entire time,

Speaker4: [00:05:22] But they're working out.

Harpreet: [00:05:24] No, it doesn't sound that that great.

Speaker4: [00:05:27] But yeah, it sounds like a tin cans connected by strings.

Harpreet: [00:05:32] Yeah. But it is all good, Eric, thank you for being here. You're one of the OGS, I think you probably

Speaker2: [00:05:37] You know you.

Harpreet: [00:05:38] Austin Ash and I know you guys were. And we go to you guys are one of the early, early

Speaker2: [00:05:45] Like, you know, adopters of

Harpreet: [00:05:46] This officer. Thank you guys so much, you guys. You guys were here back before I rebranded it to Happy Hour. I rebranded it to a happy hour just so I have an excuse to drink beer at 4:30 in the afternoon.

Speaker2: [00:05:58] But, but [00:06:00] yeah, do thank you

Harpreet: [00:06:01] So much, Eric. All right.

Speaker2: [00:06:02] So maybe once

Harpreet: [00:06:03] We get your microphone situation figured out, we'll hear what you got planned for the next two weeks. Also, thanks now. Yes, so much better.

Speaker4: [00:06:10] Oh OK. I keep hitting my chord and like, rips it out, OK?

Speaker2: [00:06:14] Anyway, next fifty two

Speaker4: [00:06:15] Weeks is just like getting more solid on my

Speaker2: [00:06:19] Professional brand because my personal

Speaker4: [00:06:21] Brand is there. I want to get solid on the professional brand

Speaker2: [00:06:24] And not

Speaker3: [00:06:25] Because I want to use it like make money and anything like

Speaker4: [00:06:28] That, but just because I want to feel like

Speaker2: [00:06:29] I have given myself

Speaker3: [00:06:31] The direction for it, you know, and just because

Speaker4: [00:06:33] That just feels good to be like setting some of those things moving forward. That's it.

Harpreet: [00:06:39] E Data, Science Community Content Creators Award

Speaker2: [00:06:44] Favorite LinkedIn

Harpreet: [00:06:46] Profile, favorite LinkedIn person, Eric, thank you so much. Also made a huge shout out Greg Kookie of Greg's been, you know, even here consistently since the beginning. Thank you so much, Greg. Joe as well, you guys Russell as well. You guys, everyone like I know how valuable everyone's time is, but the fact that you guys just carve out time to come and help people like it means a lot. To me, this thing would not be what it is without all of you OGS in the field, dropping, dropping knowledge and what's going on, man. How far you come in the last fifty. Where are you going in the next 52?

Speaker5: [00:07:19] At first, congrats on a full year of the happy hours. This is incredible. I'm so happy it keeps going. I look forward to these every week when I can make it in

Speaker2: [00:07:28] In

Speaker5: [00:07:29] Terms of what's happened to me in the last 52 weeks.

Speaker2: [00:07:34] Honestly, a lot of it is

Speaker5: [00:07:36] Interpersonal and related to this. I've met probably at least 50 to one new person a week

Speaker2: [00:07:44] That I've had a

Speaker5: [00:07:45] Meaningful connection with through

Speaker2: [00:07:46] My podcast. A lot of the people

Speaker5: [00:07:48] Who I've met here. And to me, that's one of the most valuable things that

Speaker2: [00:07:52] That I've gained is that

Speaker5: [00:07:54] I've been able to have really deep, rich relationships with people through this community [00:08:00] that that I didn't have previous to this from a more like outside of that, from a quantifiable perspective. You look at LinkedIn, you know, I probably had very little presence there, maybe less than a couple of thousand people relationships there to now over something 50, 60 thousand or whatever it is. My audience is on. Different platforms have also grown, you know? Quite a bit in that range, and hopefully that's because I've been producing good

Speaker2: [00:08:27] Things, I think, over

Speaker5: [00:08:28] The last 52 weeks. I've produced hundreds of pieces of content, which is something that I'm personally very proud of, and those are the things that I celebrate more than any any of the other growth because those are the measurable things that I can control. What's next? I'm still working on that right now. I'm taking a break after my really busy last month or two, but I'm excited to start creating more content again, and I think I'm really happy when I have routine and consistency. If I'm creating videos,

Speaker2: [00:09:00] I'm meeting new people,

Speaker5: [00:09:01] I'm producing content, I'm having fun. All of the other positive things take care of themselves. So that's kind of my hope for the next 50 to is. I can keep your routine and keep things plugging along

Speaker2: [00:09:13] In the same way. And again,

Speaker5: [00:09:14] Just to remain grateful for for everyone here and the relationships that I'll continue to make.

Harpreet: [00:09:19] Absolutely love it. If you guys have not been following Ken's Instagram Stories, he's been on an adventure to the last few days getting the VIP car treatment and and hotel room treatment hopefully had a good time out there, man. Yeah, do definitely relationships, man. Some of these relationships that

Speaker2: [00:09:35] Are formed have quite literally

Harpreet: [00:09:37] Changed my life, like the opportunities that have happened

Speaker2: [00:09:41] To just connect with, like OGS

Harpreet: [00:09:43] In the game, like Joe and Tom and Vin and Greg, and just all these amazing folks.

Speaker4: [00:09:48] Ben as well. And.

Speaker2: [00:09:50] Not just the meeting, new people,

Harpreet: [00:09:53] The fact that.

Speaker2: [00:09:55] Ok. I got a

Harpreet: [00:09:55] Job from doing this. It's pretty crazy. Let's uh, yeah, [00:10:00] let's go to what did Deepak, what is he still here, Deepak, man? Well, I don't think I've ever seen you in the happy hour, but this was one hell of a happy hour to come to its one year anniversary of this man.

Speaker2: [00:10:10] How are you doing?

Harpreet: [00:10:12] How's your last 52 weeks been? What's the next 52 weeks going,

Speaker2: [00:10:15] By the way? Shout out to the new ternary Data.

Harpreet: [00:10:19] I probably butchered that name, Joe.

Speaker2: [00:10:21] You got it right.

Harpreet: [00:10:23] Data the folks that that are in the building. Aaron and Matt and

Speaker4: [00:10:28] Whose lives

Harpreet: [00:10:29] Lost. But I think come

Speaker2: [00:10:31] In for

Harpreet: [00:10:32] Everybody that is tuning in. Let me know if you got questions. If you're in the room, let me know if you got questions. I'll call you up if you are watching on LinkedIn or on YouTube. Also, I will cue you up Tor. What's up? I see you on LinkedIn man myth having you in the room.

Speaker2: [00:10:46] Thank you for being such a supporter.

Harpreet: [00:10:49] If anybody doesn't have questions, we are

Speaker2: [00:10:51] Going to just keep on

Harpreet: [00:10:52] Going through

Speaker2: [00:10:53] With these. You know, tell me

Harpreet: [00:10:54] How you've been in the last fifty two. Where are you going the next 52? Also shout out to Vivian Vivian, I see you. Thank you for being here. All right, Deepak, hopefully I killed enough time for you to have a response. It looks like your microphone is just looks like your microphone is. Oh hey, you got it.

Speaker4: [00:11:11] Yep. Yeah, I'm Deepak and I'm new and it's my first week off. Like, like, I just attended these sessions. Like, I'm interested in data analytics and every day like learn like I did my master's in analytics from Texas A&M back in 2017, and I went back to Nepal and

Speaker2: [00:11:32] I started working.

Speaker4: [00:11:34] And now I'm in New Zealand, like working as a strategy and planning analyst for like City Council. Like, I started this role just like on 29 September a few days ago, and it's my best to do learning and analytics like and I

Speaker2: [00:11:53] Follow people

Speaker4: [00:11:54] From analytics and everything that I like. I look into like any kind of learning [00:12:00] materials, and I follow people and I feel like in this place, like I can learn like like I can increase my personal branding professional branding too, and I can learn different things from different parts of the world. And thank you for creating this platform and congratulations for like one year.

Harpreet: [00:12:23] Yeah, man. One, you're in the game. Thank you so much. Shout out to to everybody else that's just winning. I see Greg has a question. So we'll go to Greg after this. But Deepak, congratulations on the new role. I'm happy you're here. Looking forward to seeing where you're going to go over the next fifty two weeks.

Speaker4: [00:12:40] Thank you so much.

Harpreet: [00:12:41] Shout out to A. AIs, the aunties in the building, and he is probably the only person that's listening to every single recording that I have pushed out all 210 hours. Shout out to Tandy for that. Also, Ashton just landed a new job about a month ago, so congrats to you for that action. Definitely looking forward to hearing from you.

Speaker2: [00:13:00] Let's go to Greg's question.

Harpreet: [00:13:03] By the way, if you're listening on LinkedIn, go ahead and smash that share button and share this with your network. Tag everybody in this room that you see. Let's get this thing going on. If you have questions, you can leave them in the chat or in the comic box. Greg, what is your question, my friend? Go for it.

Speaker4: [00:13:18] Yeah, yeah, yeah. First of all, Harp, kudos to you, man. And happy anniversary to the office hours. And you know, this is just a translation of two key things that that you know anyone can have or to achieve success. And you, you have them, you have consistency and you have determination. So you show up every Friday, but also kudos to everyone who show up on a regular basis as well. I'm always on the learning, you know, mode whenever [00:14:00] I come here and it's it's amazing to listen to you all. So my question is about I was reading this article the other day and I and I never thought about this. So throughout a data science lifecycle project lifecycle, you know, you come up with a business problem, you do all the cleaning, et cetera. You develop a couple models right to make sure you do some cross validations and then you may find that three of them may be good, like give you good results and you may do some sort of ensemble and find out based on some ensemble of these three models give you even a better result than having them individually. Now is there a way? Is it like, um, good practice to even decide to put that in to production? If so, what are the what are the caveat to doing that? Like, what are the things that would make you regret pushing something like that to production if you cannot find something better than just that ensemble?

Harpreet: [00:15:08] Yeah, so I've definitely done that, I've I've once ensemble five models together and spit that out as prediction, it becomes mostly an engineering challenge. I mean,

Speaker2: [00:15:20] You still have just one

Harpreet: [00:15:21] Single API where the data comes in, but hopefully all of your different algorithms are, you know, predicting off the same Data model. I mean, that might be needed that way, but let's hear from from other folks. Can I see you have your hand up and I'd love to hear from either Joe or Matt Housley after that can't go for it.

Speaker5: [00:15:44] Yeah, this might be on the more obvious side, but I find myself also just forgetting this from time to time is that like model complexity can lead to like the more complex models are, the more things that could possibly go wrong. I think we're [00:16:00] all familiar with Occam's Razor and

Speaker2: [00:16:03] Just being careful

Speaker5: [00:16:04] About OK, if we have whether it's more data sources, more different algorithms we're using, there's more considerations we have to take into play into how data can be skewed or in like, you know, whatever our different algorithms are optimizing them. So I don't think that that's a like as actionable a thing about putting that in production, but it is something we should always be thinking about. Is is the other model so much more simple with very similar results that might be a still a better one to go through because it's easier to explain or it's easier to understand, or we can see that the faults in it. So that would be my my two cents on that front.

Harpreet: [00:16:43] Yeah. And actually, that's an excellent point. So when I did deploy those five models to production and averaged out the prediction, what I actually ended up, what I was doing on the back end drag was

Speaker2: [00:16:53] Actually was trying to figure

Harpreet: [00:16:54] Out which of these five models actually performs better on unseen data because during training, they were all getting really close. Like like statistically, the predictions were not off by. Much like I couldn't statistically say that one of these models was better than the other. So I said, let me push them all to production, average the prediction and, you know, collect more data. And then again, we're going to test to see which one ends up doing better

Speaker2: [00:17:19] In unseen Data and then

Harpreet: [00:17:21] Promote that one as the only one that serves a prediction. So that was kind of pipeline I was going for and I did do that five model deployment. But yeah, it's definitely that that's what we would continue doing. Let's hear from,

Speaker2: [00:17:33] I think, Joe next, Joe or Matt. Yeah.

Speaker4: [00:17:38] Oh, well, no. I mean, ensemble learning. It's interesting. One of the first

Speaker2: [00:17:44] Things I did with

Speaker4: [00:17:45] Machine learning way back in the day was. See the ensemble learning with the Osbournes as before the days of a

Speaker2: [00:17:53] Deep learning, deep learning,

Speaker4: [00:17:55] So. I think it's a good approach. I don't see much wrong with [00:18:00] it in production, I would say you just got to make

Speaker2: [00:18:02] Sure you're not overfitting.

Speaker4: [00:18:04] Typical things you do want to do is obviously make sure you're monitoring for the time you have to retrain your model. So I don't see any issue with using ensembles I used quite often for quite a while. Erin, do you have any thoughts on ensemble that much with that? I mean, explainability is a big deal, and I don't know. I mean, how accurate I liked what Ken said. How accurate doesn't need to be. You're getting a few more points where you could use a simpler model and then what are you? What's the production use case? Is it like an overt product where you're doing like a recommendation engine or a self-driving car, something where you know it's got to be fast and perform or you ship enough weights for, like other people to consume and use what feed into that? How we when we used it at Overstock, we sold furniture, so a few points doesn't really help a whole lot with selling furniture versus the complexity of managing it, so I'm always going to break it down to how do you manage the model in the lifecycle? And what's the cost of keeping that in production versus a few points in accuracy? If you're doing medical equipment, then probably that few points is worth it.

Speaker2: [00:19:28] I don't

Speaker4: [00:19:28] Know so. Since you kind of called me out, that's my response. No worries, man. Yeah. Can you guys hear me OK? I'll give my advice on this. I'm sitting in the courtyard. There might be music for that perfect, perfectly clear. Yes. So one thing I like to emphasize is just get a decent like, very basic model out quickly to the point where I think a lot of data scientists like to be really critical of prepackaged models and platforms. My attitude is get something out the door, so it works, and then you can start running your refinements kind of behind the scenes pre-production. [00:20:00] You've got them running, you got them automatically training. You can measure their results versus actual results and then deploy more advanced enhancements over time. I guess my advice on ensemble learning would be a refinement. Get something simple, practical working out the door and then gradually, maybe you can add more things to ensemble if you have any measurable differences. Yeah, the issue with Ensemble two is every time you have to retrain, you're probably going to get a different ensemble. That is how the cookie crumbles on that one, so. It's good, though, we should use it.

Harpreet: [00:20:35] How many anything to add there and apparently Mexico is in the building, but I don't see it in Mexico. Shout out to Mexico,

Speaker2: [00:20:41] One of the

Harpreet: [00:20:42] Ogs at the office hours.

Speaker2: [00:20:43] Thank you for being

Harpreet: [00:20:44] Ever so helpful. Also shout out to Dylan and Christine, who have not seen in a very, very long time. Get to see back, my friends, Mexico. Go for it. Do you hear? I see your hands raised.

Speaker3: [00:20:56] All right. The really fun part is I accidentally hit that button. I wasn't trying to raise my hand. All right. Yeah, but actually, I'm trying to. I'm trying to find article by Chip Quinn, who wrote, It's something along the lines of data. Scientists should not have to know Kubernetes, which which is necessarily her central. Ok, so that was the central point. But as part of that article, she talks about how there's a lot of like leaky abstractions between the dev and prod environment. And so from like the email and Gmail ops survey team perspective, the way we kind of treat it is like, can you package it up as a Python package or could you can you put it behind an API endpoint? And that's kind of like the sort of bias, which is that to some degree, like for data science teams, they're sort of focused on the experimental like dev environment, so they can kind of experiment with different models, whether it's ensemble [00:22:00] where, whether it's deep learning for us on the email, in general, upside, we're kind of like, OK, well, can we? Integrated into a pipeline and infrastructure. So that's kind of like how we sort of view it in the sense of. I don't want to say it's it doesn't

Speaker2: [00:22:18] Matter whether it's an ensemble, it's

Speaker3: [00:22:20] An ensemble model, but essentially. To some degree, like. In the longer term, the longer scheme of things, sometimes just having a really simple model that works really well, that we can monitor, that we can measure the performance of. It will go a much longer way than something that is architecturally very, very complex and for which we may not be able to experiment or measure the performance of or a B test.

Speaker2: [00:22:49] So that's kind of like

Speaker3: [00:22:50] My two cents. But yeah, sorry, the raise hand was accidental.

Harpreet: [00:22:55] Well, it was good to see you here again after so long.

Speaker2: [00:22:58] Presence was missed. Thank you for the

Harpreet: [00:23:01] Insightful comment there. Also, shout out to to A. Who is in the sauna? That's awesome. The sauna is an amazing place. Let's go to mark who says wouldn't it be more dependent on the DAG? Talk to us about that. Then, after Mark will go to Matt Blaze, I

Speaker2: [00:23:14] Just thought, You raise your

Harpreet: [00:23:15] Hand? Matt. So let's do this. Let's actually go to Matt, and then we'll go to Mark.

Speaker2: [00:23:21] Yeah. So my work is

Speaker3: [00:23:23] Starting to use something like like Auto

Speaker4: [00:23:26] Automl, they're using Ml Flow. Does anyone have like any advice on like on that or any other systems that are similar to ML flow? I know how email flow works.

Speaker2: [00:23:36] It's like recording

Speaker4: [00:23:37] The experiments, the artifacts,

Speaker3: [00:23:40] But

Speaker2: [00:23:41] I'm just I'm very new to it, so I'm just

Speaker4: [00:23:43] Wondering if anyone has any advice or any insights on that. Yes. For that,

Harpreet: [00:23:47] You can use Comet four. That comet is an amazing platform for you to manage all of

Speaker2: [00:23:53] The experiments, and it's

Harpreet: [00:23:54] Free. So tiny ELLE.com for a comet, check out comet. [00:24:00] Seriously, though, it's an epic

Speaker2: [00:24:02] Product and it is.

Harpreet: [00:24:05] It's useful as hell. So probably comet. Mark, let's go to A.. Let's go to to your question. Otherwise, you all start talking about weights and bikes and stuff that could be a conflict of interest in me. It's just a comet, right? Comet. Mark, where are you? Are you still here? Don't see Mark. All right, looks like Mark has dropped from the call, let's go to him. Go to Austin's question, Austin. Go for it.

Speaker4: [00:24:35] Yeah, so my question is just around the when trying to just kind of take in the landscape of what's going on in the field, whether it be focused like you're focused on a particular topic for the

Speaker2: [00:24:50] Moment or you're just trying to

Speaker4: [00:24:51] Keep up on, oh, here's something new that's coming out, whether it be a library software

Speaker2: [00:24:57] Or something.

Speaker4: [00:24:58] What kind of strategies do people employ to smartly intake that information? And kind of how do you filter it down? Because I realize I, I probably should do a little bit more reading and I want, but I want to be more focused and efficient about doing it. So what are some strategies that people use to kind of. A filter that down when there's just so much out there.

Harpreet: [00:25:25] Yeah, let's go to let's go to Vivian. Then after Vivian, let's go to Dylan. Vivian look surprised. I didn't mean to catch you off guard.

Speaker4: [00:25:33] Yeah. Can I not go first? I'm sorry.

Harpreet: [00:25:35] Ok. Yeah, definitely. Let's go to let's go to Dylan.

Speaker4: [00:25:39] I was about to say the same thing, could you repeat the question? Yeah, so when she trying to learn a little bit more about what's going on in

Speaker2: [00:25:48] The field, AI,

Speaker4: [00:25:49] Machine learning, Data, whatever. How do you go

Speaker2: [00:25:52] About absorbing and being more

Speaker4: [00:25:56] Focus because there's there's so much available out there. What do you what kind of strategies [00:26:00] do you do to like find focus

Speaker2: [00:26:02] Things, but outside of

Speaker4: [00:26:03] Maybe doing towards data science or analytics? Video Are there other areas where you go to to get that information? Yeah, I kind of took a couple pronged approach. One with books just started with one and then as references were added in there, kind of checked those out. Another part, different newsletters and just kind of picking very narrow ones like the Andre Birkoff. One has always been great for me and kind of just building off of that. Just kind of taking it as much as I can and then podcast as well. Harp rates has been great. There's been a bunch of episodes on there that have actually gone further and checked out those people, and that's been really helpful when you get like a one hour preview of someone and how they communicate. And if you like the information that they have, most people that are on a podcast are going to have further information out there, whether it's their website, LinkedIn, whatever else. And then just kind of following down that rabbit hole and seeing if it gets me where I want to go. But we will say, for me, it's been a lot of trial and error. Some of them don't lead great places, some of them the information isn't as great as you want, and some of them people just aren't as knowledgeable as you hope.

Speaker2: [00:27:12] Yeah, hopefully

Speaker4: [00:27:13] That helps a bit.

Harpreet: [00:27:15] Thank you very much, Dylan, I see Joe and Mark had their hands up. Let's go to Joe. Then after Joe will go to Mark and then then I'd love to hear from hopefully Vivian or anybody else that wants to chime in. If you if you got any tips for our friend Austin. Go ahead and smash the Reactions button. Raise your hand. Let me know. So that way I can add you to the queue. Joe, go for it.

Speaker4: [00:27:38] It's insane. Who saw Matt Turk's Data landscape infographic the other day?

Harpreet: [00:27:44] Yes. So you posted that in the MLS community slack? Yeah. Yeah, it's like,

Speaker4: [00:27:48] I think, more Data startups than there are atoms in the known universe right now, so it's kind of crazy. There's a lot to know and it keeps growing. So I mean, the approach is I [00:28:00] tend to kind of second guess how I would approach things now. But the way I have done it for the longest time, I have an iPad and I just queue up articles throughout the week onto it to the point where there's. I mean, anywhere from 30 to 100 articles, and I spend the weekend just reading those lists, like reading habits are pretty legendary. So there's no shortcut, though. I mean, you got to filter through. I think it just read a lot of stuff, and I get a lot of stuff from maybe Hacker News newsletters and, you know, just random stuff I find online. And I think between all that, that tends to be a good way. I tend to read very widely too, because I think there's a lot of things. If you focus too much on data science property, you tend to miss the bigger picture of what's happening in other fields that will eventually influence Data science, for example. So I read a lot of programing blogs. I think a lot of things start as software engineering and usually filter down from there because it does tend to be ahead of the curve in terms of practices. So that's some suggestions. Youtube's also great books are great, but but I would say the challenge of books is you've really got to pick your books carefully because by the time they come out, you've got to realize that that writing and editing and release process is quite lengthy. And at the rate the Data fields changing right now, there's no guarantee you're going to get current information making process focus on the fundamentals or specific topics. That's also a great approach as well. There's a lot of noise out there, but you get signal that focusing on fundamentals. So that's my two cents.

Speaker2: [00:29:32] So thank you very much.

Harpreet: [00:29:34] So shout out to some new faces I just saw real quick before you get to market or just to say what's up to Tegina and Thomas? I think it's Thomas. Also shout out to Al Bellamy. Good to see all you guys here, Matthew, because I see you guys here. And by the

Speaker2: [00:29:47] Way, real quick

Harpreet: [00:29:49] Shout out to to next thing next.

Speaker2: [00:29:51] Going on, man. Happy to have you here.

Harpreet: [00:29:54] So let's go to Marc Freeman and then after that, we'll go to Eric [00:30:00] and then coast up you guys managing this learning thing, man. All the shit out there for you guys to learn a lot of stuff to wade through. How do you figure out where to go, what move to make? By the way, if you guys got questions, let me know. I'll add you think you can drop in the comments or in the chat right here? Marc, go for it.

Speaker4: [00:30:20] Everyone so similar to other people in newsletters are are amazing because they have someone's already curated for you. One of my favorite is Data Elixir. That's one of my favorite ones to read. And also Andrew's deep learning, or TBI newsletter has been really good as well. But beyond that, something I really like her company blogs and the way you kind of think they're in is like in the Ml space. No one's really figured out how to do it end to end well. And so you have all these kind of piecemeal pieces of infrastructure. So like, for example, Commente has like the the the experimenting management component, you know, five champion have like Data Data ingestion component, right? So there's different pieces of the chain. And so where are you interested in that chain? Are you in the analytics side or are you in the projection side? Maybe some data ingestion side identify what the top companies are and then identify their blogs and just read through the blogs. And the key thing to remember is, like one of those bloggers are trying to sell you something, so take it with a grain of salt, but you're seeing like, what's their framework for understanding a problem? And they're trying to sell to a particular use case and and like have a particular value proposition to a problem in the market, which is probably something that's larger in the greater market as well. And so those are similar problems you can solve. And so like, how are they approaching different problems for that? And so. My favorite blogs are like the Databricks blogs. I really like just reading how they're thinking about data for things. Again, they're trying to sell you something, so take it with a grain of salt. But I think just reading through how how [00:32:00] different companies are also like through podcasts or interviews, how they're talking about the problems they're facing and how they're trying to solve it has been really helpful for me to learn about the space and like what's new out there and also like how to apply my skills

Harpreet: [00:32:11] And to sell you some everyone, even me. I'm trying to sell you me to it. Let's go. Let's go to Eric and then coast up and then Nick Yo, if you want to chime in here, I'd love to hear from Nick. If you guys don't know Nick, he just released a best selling book, so. You can tell us a little bit about that, as well as Ace, the Data size interview.

Speaker4: [00:32:34] I'm here to sell something, you know, he's not wrong. I'm here to show support.

Harpreet: [00:32:41] Yeah, let's go to Eric, then Costa, then Nick. And then if you guys have questions again, like just let me know in the chat, I actually think you

Speaker2: [00:32:48] Were in the comment section if you want to

Harpreet: [00:32:50] Chime in on this. I mean, I think Ken probably would have some good, good words of wisdom here as well as I'd love to hear from you. Go for it, Eric.

Speaker4: [00:32:58] Yeah. So kind of similarly along the lines of what I was saying about sticking with the fundamentals, and that's to filter out this the extra stuff. I just want to know what's going to be useful to me. And so if I am looking to work with a certain person or work with a certain company or whatever, it's like, Well, I'm just going to go and I'm going to look at that thing or look at what they're interested in or

Speaker3: [00:33:24] What their needs are.

Speaker4: [00:33:25] And then that helps me decide where I want to start. And I'm, you know, dumb enough in enough ways that starting with the fundamentals is where I'm pretty much always going to start. And so and so that and that that helps me also like totally relieves a ton of pressure because what I'm looking for a lot

Speaker2: [00:33:43] Of times is I'm looking for

Speaker4: [00:33:45] Validation that what I want to learn or what I think is

Speaker2: [00:33:47] Interesting is valuable

Speaker4: [00:33:49] And enough

Speaker2: [00:33:50] That it's sufficient because if I can see that somebody that I look up to or appreciate

Speaker4: [00:33:57] Also thinks that something is valuable, well, thank goodness I don't have to go [00:34:00] and try and learn everything

Speaker2: [00:34:01] Else as well because

Speaker4: [00:34:03] Nobody got time for that. And so that's kind of how I decide what I want to do next and sift through it.

Harpreet: [00:34:09] Awesome. Great advice, Eric. Let's go to the coast up next and then go to Vivian stuff because that was Vivian than anybody else when she me, let me know. Definitely to hear from Nick and Ken as well. But of course, that will be Vivian.

Speaker4: [00:34:24] So this, to me, is a classic robotics problem, right? It's an exploration versus exploitation kind of situation. You got so much stuff

Speaker2: [00:34:32] Out in the world to learn.

Speaker4: [00:34:34] You can try to learn everything. But good luck with that, right? You can try to figure out, Oh, what's the latest stuff? But good luck with that. By the time you find and verify that it's the latest, someone else has published something new in, you know, mind blowing on it. You can try to figure out, Oh, is this the perfect solution? Like, This is the perfect thing that I need to learn next and still waste your time trying to figure that out rather than actually learning it. So I'm trying to kind of change steps now and just pick up anything that passes the the basic pub test, right? Like I have a look at like you said, Eric, like other a few people that have recommended it is, you know, some reasonably well known researchers behind it. For example, what's the sources from and

Harpreet: [00:35:15] Just kind of filter a little bit by that?

Speaker4: [00:35:17] And then it doesn't really matter if I'm learning specifically the best or the latest stuff. The fact is, if I haven't read it, I'm probably going to learn something that I didn't know before, right? That's one of the biggest things that I'm trying to change in my mindset before I was like, Oh, I need to make sure that I plan out my learning so that it's perfect. I just don't want to waste the time doing that anymore. The other side of it is also we're trying to be experts across a very broad, a very broad field, right? I might question that. I don't think we all can be experts across the whole broad, wide, amazing field of data science, right? Me personally, I'm focusing on mostly vision problems, and I pick up a few things here and there [00:36:00] about non-visual related problems. But if I focus myself on the vision related problems because my interest comes from a robotics background, I can then, you know, I can then develop expertize on that front and then you can team up with other people that don't know and then you learn way faster in that way. If we're all trying to be, I mean, you need a good balance of generalists and specialists in any field, right? So I think you're naturally seeing people. Some people exploit a deep set of knowledge and other people broaden up a little bit more. So these conversations kind of bring

Speaker2: [00:36:29] That mix together. Also, thank you very much. Yeah, computer vision

Harpreet: [00:36:33] Is interesting, I wonder, can we teach computer sight the perception computer foresight? I don't know, man. That's interesting philosophical question. Let's go to let's go to Vivian that Vivian. Let's go to Ken and the mosquito.

Speaker2: [00:36:44] And then and then Nick. Yes, I told

Harpreet: [00:36:47] Said, I'll go to knickknack. By the way, if you guys got questions, let me know I'll go ahead and add you to the queue. So far, I've got Greg up for another question. And then after Greg, I got Ken up for a question on Ml Ops. Vivian, go for it.

Speaker3: [00:37:01] Um, well, I guess that a lot of people kind of like iterated things, I was thinking about stuff so. Just one thought I had to add sort of through this pile of wisdom is that I don't know. Like when I first started learning Data science, I felt like I was trying to, like, consume the internet basically as a whole. And it was really overwhelming. And I still find myself like falling into that hole sometimes and feeling like I have to read it all and do it all and consume it all. And instead, I have specifically just picked a few things that I like and backed off and then just trusted that like the right information will come to me at the right time. If I'm like following my own curiosity and interest and stuff like that, like, I don't know, I don't know. I just feel like it can be really overwhelming

Speaker2: [00:37:50] Trying to be on the

Speaker3: [00:37:52] Forefront of everything, kind of like what people have been saying that it's so hard. So instead, just like lead with your curiosity and like, [00:38:00] read what feels interesting and then just trust that the right information will find you at the right time. And it's worked so far for me. So.

Harpreet: [00:38:11] Erin, thank you so much. Let's go to Ken then than Mexico, then Nick.

Speaker5: [00:38:17] Actually, Vivian, that is extremely relevant, that is I think what I'm going to describe, what I do is like a very natural extension to what you do. It might be also just like a completely ludicrous thing to do, but it's worked OK, well for me so far. I have like a separate YouTube channel. I have a separate Twitter feed. I have separate Instagram account where I follow very specific things and I only watch videos that are in line with things that I want to learn. And the algorithms that they use are pretty good on those

Speaker2: [00:38:46] Platforms, and they keep

Speaker5: [00:38:47] Recommending

Speaker2: [00:38:48] Me things that are in line

Speaker5: [00:38:49] With what I want to learn. Like, if you're using these platforms and algorithms

Speaker2: [00:38:53] In a way that it's

Speaker5: [00:38:55] Designed to recommend you things that you might be interested in. That's a very effective way to use them for learning and picking up new things. I mean, there are obviously downsides of that. Like, maybe it's an echo chamber for what I'm seeing on YouTube or something along those lines, but I am inherently lazy in the in the search for for that type of knowledge and setting up systems or using existing smart systems that are in place to feed me information that I'm likely going to be interested in or could be cutting edge or whatever it might be is something that I found is a little bit of a mind hack, a life hack. So I don't have to. Do what everyone is describing in terms of that, like overwhelming. Like volume of things out there to pursue. So again, I don't know if that'll help anyone, but it has been pretty interesting for me and I think fairly effective to set it up like that.

Harpreet: [00:39:49] And thank you so much. So apparently, Kenji has a finsta. So let's go and try to find

Speaker2: [00:39:54] That that fence

Harpreet: [00:39:55] That I can has Makiko go for it then? From Akiko, go to Nick. For Nic, we'll go to

Speaker2: [00:39:59] Joe [00:40:00] and then

Harpreet: [00:40:00] Ashton, you got a great comment. I would love for you to to unleash that comment from the comment box and read it out for us.

Speaker3: [00:40:10] And then with the cultural references, timely cultural references.

Harpreet: [00:40:14] So don't let the gray hairs

Speaker2: [00:40:15] Bleed until still hip

Harpreet: [00:40:17] To the streets.

Speaker3: [00:40:20] It's funny the I feel like the learning, the like, how does what is the best way for one to learn given their sort of career goals?

Speaker2: [00:40:30] Like, I've been having some

Speaker3: [00:40:31] Of these discussions

Speaker2: [00:40:32] Like at work, because

Speaker3: [00:40:33] What I've noticed is that the team that I'm on, we have like a really wide variety of skills that we can contribute to solving some really interesting challenges. So we have staff senior non senior male engineers and some people tend to aside from having a deep number of years of experience. Some people prefer to work on more, for example, technical tooling projects, whereas other people prefer the more sort of process oriented

Speaker2: [00:41:02] Like, let's just get let's get the ball to production.

Speaker3: [00:41:05] Let's interface with the model monitoring team. Let's interface with Data engineering. Whereas other people are like, Hey, we want to build

Speaker2: [00:41:12] Internally internal tooling

Speaker3: [00:41:13] To help facilitate that. And so I was having this conversation with my director because I'm like, I. I kind of want to figure out what my next steps are.

Speaker2: [00:41:22] And she

Speaker3: [00:41:23] Said that back when she was over at under armor, the way they kind of thought about their careers and consequently learning was

Speaker2: [00:41:32] Three

Speaker3: [00:41:33] Personas. One was the technical specialist persona. The second persona was like the tech lead, and the third persona was the technical strategist. Because I was her like, I don't know if I have the I don't know if I'm people becoming tech, even though I really want to do that kind of work. And when she laid that framework out, I thought to me and made the most sense because for the technical

Speaker2: [00:41:57] Specialists and it's not

Speaker3: [00:41:59] A seniority [00:42:00] thing necessarily, it's that some people are very, very much so interested in how do we take existing tooling? How do we architect it? How do we design it such that we drive business value, whereas other people are much more interested in

Speaker2: [00:42:12] Developing tooling

Speaker3: [00:42:13] And they tend to work more on what's considered like the platform side? So I think part of that also. What is your sort of career goal? What was the sort of north star of the kind of work you want to be doing? I feel like that also kind of influences what one's sort of learning pattern would be and how they would

Speaker2: [00:42:32] Appreciate, like

Speaker3: [00:42:33] How they would approach, like study, research, et cetera. Like, for example, I don't look at any of the deep learning newsletters anymore because the reality is that most of our data scientists are either like, we've set up tooling such that they're either packaging all their models as like libraries that we use internally or we implement

Speaker2: [00:42:52] Them as like runtime

Speaker3: [00:42:53] Services, whatever. But the way we interact with models is we look at code reviews, we look at how they run the tests, we look at how it integrates. So for me, it's like there's no benefit in a way to diving really, really deep into any particular technology. But there's a lot of benefit to me understanding the patterns that specific technologies use

Speaker2: [00:43:16] And how they kind of work with each other.

Speaker3: [00:43:18] So I don't even look at any of the deep learning newsletters or anything, but I keep very much so up to date with, like

Speaker2: [00:43:23] The community newsletter.

Speaker3: [00:43:25] Whereas there are people on the team who are like, definitely technical specialists, they're really thinking about like, how do we build this tooling? So they're diving way deep into, you know, like a python bites or they're looking at Python or whatever. But I look at that too, right? So I think that is something that is kind of really important is that when you're thinking about your learning sort of like mode, it's figuring out what you sort of want to get out of it. Because if you're really technical strategists, it probably would not benefit one to dove deeply into like everything and vice versa if someone [00:44:00] is a technical strategist. It would be good to understand like what's going on the like, what are the trends, the patterns. But they might benefit more from reading white papers or research papers like algorithmic development or things like that. So.

Speaker2: [00:44:16] Nico, thanks so much. Let's go to

Harpreet: [00:44:18] First, let's hear from Nick and then Ashton, then we'll go to Joe and

Speaker2: [00:44:21] Then Erin.

Harpreet: [00:44:22] Again, the topic is, look, man, there's so much out there to learn. Life is the life is short. The craft is so long. How the hell do you manage all the input? How do you figure out what it is that you should be focusing on in this field of data science? Next thing? Go for it.

Speaker4: [00:44:39] What's up? Thanks for having me on. My first time hanging out with all you guys, so this is cool for me. What I do is I intentionally, I'm someone with a bunch of different interests. What I do is I intentionally set like an hour or two in the morning to explore those interests, like scrolling through Twitter. Except my Twitter is like filled with a bunch of product and Data and B.S. and tech Twitter types. So it's like educational. It's not like meme Twitter, but I intentionally know that the rest of my day should be like the focused energy time where I'm like working on one thing or learning one specific thing. But that extra hour to that I intentionally set off is like my time to like, just like learn about anything and everything.

Speaker2: [00:45:19] And just like

Speaker4: [00:45:20] As I think Ken had mentioned, just like the feed does a pretty good job of like you tell it, like you're the 50 accounts you want to follow, and then it finds you content from them and related content. So if you just follow a bunch of Data product, B.S. and tech people, then you get a lot of good articles around that forever. So that's just kind of how I set it up. Yeah.

Harpreet: [00:45:40] Thank you so much. So let's let's go to ash before we go to Ash and shout out to George Farah Khan and George Farah Khan is in the building. Thank you so much for coming and hanging out, and

Speaker4: [00:45:49] I think you Harp we just wanted to wish you a happy anniversary. Congrats on your first year and all the amazing talks and all the amazing content that you're putting out there. So congrats and many more. And [00:46:00] the cats, as I

Harpreet: [00:46:01] Thank you, George. Thank you. Yeah, I was hanging out with George at the beginning of the month in Vancouver. Not only did he did, he managed to strategically plant his hair into my dish so that we can get get get a discount. He was generous enough to know, but the place we went to is amazing. I forget the name of it, but they took really good care of us, and George is kind enough to give me a ride to the airport afterwards. George, thanks so much. Let's go to ash and ash and then after Ash and Joe and Erin and then we got a question coming in from Greg, which will move to right after this. And then Ken has the annual OBJ question and then Gina has a question as well. Go for the action.

Speaker4: [00:46:39] Great. Yeah, congratulations on the one year and yeah, happy to be here with everyone. Yeah, lots of good points. Absolutely. Like, I won't repeat everyone's points too much, but one of the my comment was in the chat dry. You know, as we all know, do not repeat yourself. But in this case, I do. I do encourage you repeating yourself.

Speaker2: [00:47:02] By that, I mean,

Speaker4: [00:47:04] I found it really helpful to kind of repeat my goals, my learnings to other people. So I'll just grab whoever is willing to listen to me, talk about everything and just, you know, get their feedback, what they think. They think I'm doing something wrong, like, yeah, I'll take that in and kind of loop it into my learning. And it's been really, really helpful to have that

Speaker2: [00:47:29] Have that goal

Speaker4: [00:47:30] In mind because I thought when I first started, I had to just be a sponge and soak everything in. And that's not the right way to go. I mean, you do get exposed to a lot of fields and you do kind of realize what you like when you don't like. But it's very stressful, so always, always repeat yourself, remind yourself what you're doing, why you're learning, whatever it is, you're learning and yeah, just develop your skills from there. And by the way, a [00:48:00] little bit about me. I just found out about Data science in March of 2020. So I'm very, very, very new to the field. But, you know, do not like I would always suggest to not underestimate the compound effect of your learning. So like within six months, I felt comfortable

Speaker2: [00:48:18] Enough to be here

Speaker4: [00:48:20] Like, you know, like talk to go to webinars and stuff. And also,

Speaker2: [00:48:25] Within a year, I

Speaker4: [00:48:25] Was able to learn all those fancy tools like

Speaker2: [00:48:27] Snowflake

Speaker4: [00:48:29] Vr and the concepts behind them. So like looking back now, one year, six months or even three months, actually, that's plenty of time to learn a new skill and be comfortable with it and be able to talk about it.

Harpreet: [00:48:46] I shouldn't think so much, and congrats on that new job and oh yeah, that's fine, you know?

Speaker5: [00:48:52] Yeah, I didn't even say that. I mentioned that it's been four weeks since my

Speaker4: [00:48:57] New job as a senior CDP analyst, CDP being customer data platform. And I'm pretty excited to learn. I still don't know honestly what I'm doing, but I'm trying to figure it out. Be excited to learn all those like more of a business side, client side sort of thing.

Harpreet: [00:49:15] Thank you very much for coming by Ashton also. Nicole is in the building, Nicole Jan is here somewhere. Nicole, thanks so much for coming in. Hey, now, good to see you as well. She put out an amazing event yesterday. It was a lot of fun. Let's go to Joe and then Aaron, and then we're going to move on to Greg's second question. And after Greg's question, there's the question from Ken on MLPs. Then there's the question from Gina, then coast up and then Lauren got a question coming in from LinkedIn. Yeah, go ahead, Joe.

Speaker4: [00:49:42] And actually this this this might be something that I can talk about together because it's something we were actually on a hike this morning talking about, which was at ternary, for example. I think I've come to realize we do Data engineering a lot differently, and we think about Data engineering a lot differently than probably other people. [00:50:00] And for people who don't know where a Data engineering consulting firm is, it's like literally all we do all day. But the context of it is as we're as we're trying to get people up to speed for certifications and stuff, what I realize is there's actually a base level of knowledge that needs to be established. And so I've been thinking about this the last few days and how teams really need to, I think, set prerequisite knowledge for other people on the team. It's it's something I hadn't thought of things before,

Speaker2: [00:50:36] But this is

Speaker4: [00:50:37] The first time where I was like, OK, so the baseline technical knowledge that everyone needs to know in a team, whether you're seeing your junior. What does that look like? And so that's something I've been really thinking about lately. I don't

Speaker2: [00:50:50] Know, Aaron, do you have any comments on that

Speaker4: [00:50:52] Or other thoughts? Well, yeah, I mean, you're exactly right,

Speaker2: [00:50:57] This has come up.

Speaker4: [00:51:00] There's that prerequisite knowledge of you're studying snowflake, you're studying Fireball, you're studying BigQuery, but you don't know what column or databases are.

Speaker2: [00:51:09] You should probably

Speaker4: [00:51:09] Know what column or databases are. You're studying deep learning, but you've never done logistic regression. You've never done like PCR or any of the basics. There's no replacement for that, that junior statistics class in that junior algorithms and data structures class. You know, you were telling me about, you know, I was getting into column family databases with with Mexico. Or maybe it wasn't Mexico, someone else we were talking about big table. And then it led us to be tree indexing. And then and I'm trying to explain it to someone and you know how fast rights happen. And it's a linked list and they're like, What's

Speaker2: [00:51:46] The link list? And I'm like, Well,

Speaker4: [00:51:48] You got to go to that basic algorithms and data structures class that if you didn't take in college, just go, just go snatch one up on Udemy, you know what I mean or something. But I'm not saying everyone should do that. [00:52:00] But and the thing I was going to add and it was going to be real short, is I need to

Speaker2: [00:52:04] Really since since I

Speaker4: [00:52:05] Learned for a living

Speaker2: [00:52:06] Almost in Joe's

Speaker4: [00:52:08] And matter kind of my boss, I got to treat my

Speaker2: [00:52:11] Learning like an agile process.

Speaker4: [00:52:13] You know what I mean? Like, and I got to really check my intention of learning. Am I trying to learn about Gans or like deep learning or something so I could like come to this data science podcast and like show off, like how mathematically literate I am and how cool I am? Or like, do I need to learn this snowflake implementation because Joe expects me to implement it like next week? And that's like, you know, I think for me, it's come down to my attention. Like, Am I learning something so I can signal it? Or am I learning something because I actually need to like, deploy it and like, create value for a team and for people, you know?

Harpreet: [00:52:54] That said, that's clutch advice.

Speaker2: [00:52:56] Erin, thank you

Harpreet: [00:52:56] So much, and Erin also has once opened for the Wu-Tang Clan, so I

Speaker2: [00:53:03] Don't know

Speaker4: [00:53:03] Where that came from. I put that on there. You open up, you open up for all the Ghostface in everybody. Anyway, it's fun times.

Harpreet: [00:53:12] Yeah, yeah, that's clutch advice. Thank you so much. Let's let's keep it moving. We got Greg's question. Then after Greg, we got 10 and then Gina then coast up. And then if you need office hours, this is typically how it goes, man. Well, we'll just we'll just answer a question until that question is no longer the question anymore, and it's just a settled fact. So, Greg, let's go for it.

Speaker4: [00:53:36] Yeah, my question is about software engineers. Do you guys think there should be a push for that? Software engineers should learn more and more about machine learning. And if so, what happens to ML engineers or what? What do you guys see there? Because, you know, maybe pushing things to production, integrating with business [00:54:00] production systems? You know, if it's if it's a challenge, what what kind of transformation are we seeing going forward with regards to software engineers?

Harpreet: [00:54:11] Earlier, Nick, for this, Nick, you got any insight here for this man, I'd love to hear from you on this. And then after Nick, we'll go to Mexico

Speaker2: [00:54:17] And then

Harpreet: [00:54:17] If anybody else wants to chime in on this question, just go ahead and raise your hand at the top of the queue. But we'll go nick to Mexico.

Speaker4: [00:54:23] And this is embarrassing. Can you rephrase the question? Sorry. So it's about software engineers. Should there be a push for them to learn about machine learning so they can become more efficient or more effective with working with data scientists? If so, what happens to the ML engineers in that space? How is that from what are you seeing? Honestly, I think like. Maybe not push them to learn more, but

Speaker2: [00:54:50] Maybe just push them to do

Speaker4: [00:54:51] More Data engineering, right? That's that's something people don't love to do, and that could yield better results just because let the more people do, the more. And honestly, engineers like you don't have to go all the way to mill, just let them get better at Data and make sure that more people have the space to do their stuff on top of good Data and good pipelines and good tooling. So I'd say that as one thing, but then my other answer would be like, maybe a little bit. I guess I'm a big fan of like cross training. Again, I was alluding to how, like my Twitter feed is just like product VC tech and Data. It's not even just one thing. And I'm, you know, I just always think that. Getting projects out the door is often about solving technical challenges, but it's also about working with people. And one of the biggest things was working with people is being able to speak someone's language, right? So we're all speaking English, but often, you know, a PM is talking one way and then an engineer is thinking about something totally different. And one best way to get everyone on the same page is like, Well, what's the Rosetta Stone like? How do we get everyone talking on the same page? It's like, make sure your people know something about the other person's job, what their field's about and [00:56:00] what they're optimizing for.

Speaker2: [00:56:01] Because that's when I was at Facebook.

Speaker4: [00:56:03] One of the biggest things would be like engineers are optimizing for one thing, and they're being graded on one thing. Pms are being graded on another and designers are being graded on a third thing. So designers here pushing me to make it look pretty. The PM is trying to just up a number that's all there bonuses tied to a single number and then engineers are trying to improve perf, which might not tie to numbers. And suddenly we're all talking different ways because we just don't know how each other acts. So I guess in that sense, it's good to cross train. But I mean, you know, software engineer doesn't want to do email and, you know, don't don't, you know, just do whatever interests you. But I think learning about Data is always a good thing.

Speaker2: [00:56:39] Data Engineering Data Engineering would

Speaker4: [00:56:41] Be like a good middle ground.

Harpreet: [00:56:43] Nic, thanks so much. By the way, Nick's got a book out, y'all should check it out. There's a link right there on the comment ace of Data science interview. We'll be talking about this in the podcast at some point, probably November or something like that. Looking forward to that, Nick.

Speaker2: [00:56:55] Let's go to let's go to a quick, quick thing.

Speaker4: [00:56:59] It's forty four dollars in Canadian dollars, so it's like thirty thirty two in a U.S. dollar. So I just want to call that out.

Harpreet: [00:57:05] This is this is Canada. Yes.

Speaker2: [00:57:08] Things are more expensive here, but

Harpreet: [00:57:10] Salaries are oddly not. Michael, go for it.

Speaker3: [00:57:16] Yes, I think so there's kind of like three layers for which that question could operate. The first one is just should anyone learn? So what should anyone know about ML? And should people know stuff about Mel? And I think they should at a high level, just because if you are a consumer of any sort of platform or product, there's a really good chance that it uses some kind of ML modeling and as a

Speaker2: [00:57:42] As a well informed member of

Speaker3: [00:57:44] Society who uses something like Facebook or Amazon, you should probably have have some kind of high level understanding of like, how are they like, how are recommendations being recommended to you? And that's very helpful. For example, if you are like a person of color and you're trying to get like a [00:58:00] home loan, you should probably be able to understand. Like, first off, not only say, how is the company generating

Speaker2: [00:58:07] Like the the credit

Speaker3: [00:58:08] Score or the prediction or the loan mortgage interest? But what are the ways that they could be generating it for you that could be adverse to, for example, and we've seen this in like the real in real estate tech, right? A couple of companies have recently gotten into trouble because they were they were providing predictions on like the the mortgage interest rate and they were using socioeconomic status or wealth and also your neighborhood index, which unfortunately has been correlated with a red line in the past. Right. So there are just things that you, as a member of society should should know. So at that level, it does have to be the particulars, but it should be like, what are the pros and cons and like, how are companies leveraging right? And where does your Data found fit in? The second level is like, as you know, as an engineer, would it probably be good for you to know? Yes, definitely. But do you need to know? Do you need to, for example, be able to describe the implementation of an algorithm? Probably not. Like what? You should be able to understand. What's going is high level like Data goes in, Data goes out.

Speaker2: [00:59:18] What are some of the

Speaker3: [00:59:19] Problems that can result from Data and from model predictions? And how what are the challenges around it? But I don't think it's like one hundred percent necessary. And in terms of like kind of where the trends are going. A lot of roles have just been specializing in companies that are big enough. You don't, in fact, like in a big company, you in fact do not have someone who is truly doing end to end. They are not gathering the data. They're not they're not cleaning it. They're not putting it in a data warehouse or Data mesh or whatever. They're not generating the model. They're not then pushing it to production. They're not also doing like observability and monitoring and experimentation [01:00:00] and doing all the stuff and big companies. You just do not see individuals having that kind of role because it is just very,

Speaker2: [01:00:07] Very hard to do all that very well

Speaker3: [01:00:09] In a way that won't get the company sued. Because if you are a big enough company, you probably have international customers, you probably

Speaker2: [01:00:15] Have to be PR. Also smaller stuff. So.

Speaker3: [01:00:18] So the net net is that I totally agree with. Nick's point is that you should probably like understand for yourself the high level concepts like where does it fit in in the product, where where is it? And as a user is probably a lot more interesting to like focus as an engineer on the Data challenges. And there's a bunch of people on

Speaker2: [01:00:36] This call who can talk about

Speaker3: [01:00:37] That. But a lot of the challenges that come up with the ML engineers are in many ways Data challenges.

Speaker2: [01:00:42] They're not really

Speaker3: [01:00:44] Aml challenges, right? Or they're challenges of the dev versus like prod environments in the different needs of the dev versus the environment. So, you know, and as a as an email engineer, I'm not super worried honestly about if people suddenly figure out a lot of those challenges, like suddenly go up a job. Probably not, because the reality is that like if you're in a space like Data, injuring engineering animal ops,

Speaker2: [01:01:13] You

Speaker3: [01:01:13] Can take those skills and build them, and you can just sort of decide to either specialize in an area or you can move into an area that really kind of values having that sort of more holistic palette of skills. But, you know, I think Joe and Aaron can talk about how, like Data, engineering is getting in like how important it is because once again, like a lot of the challenges that we see in email ops MLA and to some degree, there challenges that have frankly already been solved in other areas like DevOps, or they are challenges that are really more relate to like Data or such things are a little bit more fundamental. So.

Harpreet: [01:01:53] Eco, thank you so much, yeah. This reminds me, I was on a call earlier today, I was talking to Dimitrios Brinkman. He's [01:02:00] the head of community at the MLCs community. If you guys haven't, you guys haven't joined that community. You should join. It's pretty active. But I was talking to him and

Speaker2: [01:02:08] He said something funny.

Harpreet: [01:02:09] He's like, bad things happen to good Data and he's going to make a shirt out of that.

Speaker2: [01:02:13] So if you know,

Harpreet: [01:02:14] If you get a shirt, that's not from him, let me somebody listen to us talking to and that idea. But bad things happen to good Data. Yeah, just remind me that as Micky was talking, let's go to Aaron after Aaron. I Mark, let's go, let's go, let's go to Marc after Erin.

Speaker4: [01:02:33] Well, I've kind of had a common theme of opportunity cost.

Speaker2: [01:02:38] If you have a principal

Speaker4: [01:02:40] Engineer at your company, let's say you're an ecommerce company and you have a principal software developer, that's a spring boot genius and all these things. Probably not the best, you know, company time to train them up in ML. I think with Mexico talking about the ethical outputs of ML, you could, you know, it's that the algorithmic Avengers and those kinds of like groups you can. I think that's that's important stuff. But I'm going to just tell you a horror story about when someone does start thinking that they know M.L.. And you know, we had we we I did product management for about a year, and I'll never do that again. And one of the reasons why I was because I understood models and I could I could talk. So they're like, Well, why don't you be a product manager and data science? And it was really ill defined, and a lot of my job was teaching the business. Ml and teaching developers ML. And you know, we we had a search algorithm we are using lter and we had some variation of that. And then once a few people learn about ML, we had a laundry list of like feature requests that people wanted to see put in our models and started getting a lot of like there's nothing more dangerous than like a really good engineer that's [01:04:00] been studying ML for two weeks and played with Jupyter Notebook and then want you to, like, make all these changes to your model because they're an engineer for 20 years and and telling me that I should add the discount to the model when we have the lowest paid price on the model and then trying to explain to them of multi color linearity and things like that.

Speaker2: [01:04:20] So, you know, there's a cost either way. And I just

Speaker4: [01:04:25] I think that that's something you want to think about. If you teach everyone about ML, be prepared for everyone to have an opinion on the model that you have in production. And I'm not saying it's a good or bad thing, but that's well, I mean, it's true. It's kind of a front end developer is expected. Everybody in the company touches the keyboard, you know, react or something, right? It's like, Yes, but I don't know, man, there's a lot of specialization. I think it's kind of funny. I actually see almost the reverse happening where there's sort of a I think Ml is still hot, but it's not like it was back in the day where it's like everybody needs to learn to code and learn, even if you're like two years old or something. Nowadays, it definitely seems like, I think now that the sort of the. You know, the announcers were off the, you know, Miles, and some some degree people are a lot more realistic about what you can and shouldn't use it for. So I think that's cool. And the nice thing is we're getting back to basics again. I think a lot of hype sort of worn off. So it's cool whether

Harpreet: [01:05:30] Or not you learned to code, you should at least learn how systems work, option, you know, how they operate, how they function. Let's go to mark. Then after Mark, we're going to shift gears, go into Ken's question for Ken. We got Gina coast up and then Lorraine's question coming in from LinkedIn.

Speaker2: [01:05:44] And I think after those

Harpreet: [01:05:45] Questions, it might be time to call it a wrap, but go for it.

Speaker4: [01:05:52] I wasn't expecting to speak, but I put a comment in the chat. Kind of an interesting use case. I'm in startup, so [01:06:00] I move really fast and many times they're like, Well, no one else is going to do it, so you do it. And so many times I'm the one sourcing the data, putting it into our database, setting up our data warehouse and then like building solutions on top of that and then putting into production. And the thing is, though, like that seems end to end, but like it's end to end like to me because I'm like, if if we're doing this like this large scale production thing, like, am I capable of doing that all really well? I don't think so. It's the fact that we're in a startup and can get away with these ones and put it out really quick that that allows that. And so the question is like, am I really doing end to

Speaker2: [01:06:39] End data science if

Speaker4: [01:06:41] It's actually just kind of like a v one to make sure the market even wants it? And so it more of a maybe a philosophical question like what's end to end? What concerns you end to end or unicorn data scientists? Because some people have made the argument that the stuff I'm doing is, but for me personally, I'm just kind of like, if you knew the kind of how it works, it feels a little hacky.

Speaker2: [01:07:05] Harp, thank you so much. All right, well, Greg, hopefully

Harpreet: [01:07:09] You have some great insights onto that

Speaker2: [01:07:10] Question. Let's go ahead and

Harpreet: [01:07:12] We're going to go to Ken's question. Guys, I'm a step away for a second because this beer is running low and I feel like we're about to go on for a while. So I've got to grab me another beer.

Speaker2: [01:07:19] But for Ken's

Harpreet: [01:07:21] Question will go to we'll go to mark, then we'll go to Mexico and then we'll go to Nick on on Ken's question. Ken, go for it.

Speaker5: [01:07:30] I think you should be actually drinking champagne today for the one year anniversary. I think you botched it, but

Harpreet: [01:07:36] This is actually a brut IPA. Oh, there you go.

Speaker4: [01:07:39] Perfect. So I

Speaker2: [01:07:42] Sat down to make a

Speaker5: [01:07:43] Video about what is Mel

Speaker2: [01:07:45] Ops the other day,

Speaker5: [01:07:47] And I read a lot of articles. I believe I know what Mel Hops is, but I couldn't define it very clearly in a couple of sentences. I think that there are a lot of components that go into it, and I was [01:08:00] wondering if anyone had a good definition from hops. Perhaps something that a child can understand or some or I could understand,

Speaker2: [01:08:08] Or it could be like

Speaker5: [01:08:10] Or if you could define it and like a like a tweet form or something like that, that was that is the challenge or the question that I would pose.

Speaker4: [01:08:19] I'll go first because I was on that list of things, so actually one of my my clients for my my side business is actually writing blog articles on maps and not necessarily like how they implemented, but more so I'm like the decision makers side of like build versus buy and like why you should consider these solutions. So it's been really top of mind if like, how do I communicate my ops and the and the various vendors in the space specifically for for like buyers? So like I think the definition that we've been going around is essentially it's like email ops and someone who's way more experienced. Please correct me if I'm wrong, because then I can write better articles. But essentially my is the ability to take your machine learning model, put it into production and one make it reliable and scalable through through various processes and tools. And so that's kind of the definition we're going around. And the key thing around that is that I talk about this a little bit earlier is that, you know, for, for me, the relatively new space and going through hyper fast. And so I think of it kind of like the Data maturity cycle where not the immaturity cycle, the the maturity cycle where initially you have a whole bunch of different players. And over time, it could consolidate. So we're like really early in that. So no one's really figured out, how do we end to end? So now you have all these piecemeal individuals saying for this, for

Speaker2: [01:09:43] This kind of process of going from like

Speaker4: [01:09:46] Data ingestion to, you know, data exploration, business use case to Jupyter notebooks and experimenting to actually putting a production in the monitoring and that cycle going back and forth. There's various players who are really good and [01:10:00] you kind of bring those all together for four ml of solution. So the MMO ops engineers are the ones who are able to really think through. You know, there's a lot of overlap between engineering and MMO ops currently. And so Michael does engineering, and she had it's really great article that they do.

Speaker2: [01:10:16] And so the

Speaker4: [01:10:17] Ais more

Speaker2: [01:10:18] So like less of the.

Speaker4: [01:10:19] All right, how I build a model versus, all right, we have this model built how we make sure it really works for our customers at scale. Every single time we can deploy it and it needs to get retrained, how can we deploy again really fast? And so that's the current thought process around it. But I'm super excited here what other people think, because this will help me in my job immensely.

Speaker5: [01:10:41] I have a kind of a follow up question to that, so for example, with like DevOps, which I believe Ml Ops is like loosely based on a huge component of that was handoff between like software engineers and

Speaker2: [01:10:57] Business people and implementation

Speaker5: [01:10:59] Or whatever it might be. How is that baked in to the system that you described?

Speaker4: [01:11:06] And we feel like, I think for one of the articles I recently wrote is like was just like talking about this use case where they they have models and the challenge that what that they describe the vendor described was essentially like typically in software

Speaker2: [01:11:22] Engineering, you're managing just

Speaker4: [01:11:23] A code base. But when it comes to engineering, you're managing a code base and a database. And that as significant complexity to that process because your code base is changing. But then also your database, the Data Data quality can be degrading if someone adds like a random value or whatnot to it.

Speaker2: [01:11:42] And so because

Speaker4: [01:11:43] Of that complexity, Emma Ops helps align on that. And so that's the explanation I was given.

Speaker2: [01:11:49] But I'm really curious,

Speaker4: [01:11:50] Like especially like I feel like Joe would have a really great idea on this space simply because all the engineering the I mean, the ops is ops [01:12:00] like we call it X

Speaker2: [01:12:01] Ops because

Speaker4: [01:12:02] Ml DevOps, Data ops, whatever. What I see the other day, Data prep ops that was one that somebody said they trademarked was hilarious.

Speaker2: [01:12:12] I talked to a guy last

Speaker4: [01:12:13] Week who can't get term DevSecOps. But really, when you have the word ops, I mean, the thing that you need to realize is if you look back to the origin dev ops, it really took its inspiration from lead. Right, so if you know what Leon is, it's a system of basically and it's a pretty low fidelity system, honestly, with the way Toyota developed it, it was like don't rely on overly complicated systems. That's actually a tentative lead you want, you want.

Speaker2: [01:12:44] But the end goal

Speaker4: [01:12:45] Is to reduce error and variation and defects, reduce lead times and that sort of thing, right? So if you apply those principles to DevOps, for example, right, that means you're automating code deployments, reducing defects in your application. I eat bugs and so forth. When you think about

Speaker2: [01:13:05] Data ops, right, you extend this a bit

Speaker4: [01:13:06] Further.

Speaker2: [01:13:06] I'm going on the food chain.

Speaker4: [01:13:08] Data means are going to reduce defects in Data Data quality issues, data governance issues, compliance issues and so forth, right? The idea is to get Data into the hands of end users as quickly as possible. Now, of course, this brings us to machine learning ops, right? And so and here's sort of the order of operations, right? You have that functioning applications. You have to have good data in order to do the machine learning, which relies upon these inputs. So machine learning ops essentially takes the same notion that remember what I said? The essence of ops is to reduce variation defects, time to value and so forth. So all you're doing in machine learning ops is exactly that. You're applying these same principles to production AIs in machine learning models. That means machine learning models being able to capture a concept Data drift quickly. You're observing defects that are predictions and you're improving the cycle [01:14:00] you're retraining. So to wrap it up, I would say when you're thinking when you see the word ops, always think in terms of I'm trying to speed things up while eliminating defects in my workflow. And that's about it, right? I would encourage you to actually

Speaker2: [01:14:17] Look at the learn about.

Speaker4: [01:14:19] There's a really good book called The Toyota Way that covers how Toyota

Speaker2: [01:14:24] Developed

Speaker4: [01:14:24] Its lean practices back in the day before lean. There was a total quality management and all these other kind of like super complicated ways of reducing factory errors and manufacturing errors. But it comes along and they're like, OK, we're just going to look at the production line and notice how quickly things move through the production line without defects. And they had things like the add on quadrants. If you just pull a quarter, there's a defect the entire production line stops, which is completely different than how things used to be. Where you just have a production line and never stop the line, that was that was horrific. Why would you ever stop a

Speaker2: [01:14:59] Line like that?

Speaker4: [01:15:00] That's that's heresy. So just a lot of, I would say, contrarian thinking that actually, it's interesting because this is we're talking about manufacturing,

Speaker2: [01:15:09] Which is like doesn't seem like it would have

Speaker4: [01:15:10] Any application to software at all. But it has everything to do with software because at the end of the day, software is nothing but pipelines and workflows. That's it. So that's my TED talk. Thanks.

Speaker2: [01:15:26] Joe, thank you. Let's go to Micki Nick.

Harpreet: [01:15:29] If Nick chime in here, then we'll go to Greg. Go for it.

Speaker3: [01:15:34] Yeah, I mean, the way I voice just kind of summarized it, it's the discipline of basically delivering.

Speaker2: [01:15:41] Sorry. My immediate or my. No, no, you get OK. It's just delivering

Speaker3: [01:15:46] Like ML products that scale resilient, scalable, all that good stuff. But you kind of wonder if they're basically like. It's not like

Speaker2: [01:15:56] Programing languages or like classes, right, like

Speaker3: [01:15:59] You can have this [01:16:00] idea of what something this aspirational idea of what something should be. And then you see, like, for example, like a dog and then you see how this dog class is implemented very differently across many, many different programs. So I kind of wonder if that's kind of one of the things we're running up against, like in the MLPs in space in that like the way EmiSwap Semillon has done at big companies is different at smaller companies and startups right before MailChimp at the start where we did like everything, including Data NML. So I was an engineer there. But this could be one of those cases where just because with startups, the resources were so constrained

Speaker2: [01:16:41] That having any

Speaker3: [01:16:43] Kind of

Speaker2: [01:16:43] Specific title would have been sort of.

Speaker3: [01:16:46] I don't want to say a lie, but maybe it would not have been quite super accurate the way that, for example, email offices

Speaker2: [01:16:54] Or Data or email and

Speaker3: [01:16:56] Just performed at bigger companies, right? Because I have bigger companies, so I take my team, to be honest, is probably a little bit more closer to ops. For one thing, I'm debugging Kubernetes and Jenkins right now. I imagine there's very little like, Yeah, I know everyone like winced at that. There's very little data science or algorithmic. It's just figuring out how the builds are done and how he can run tests for a Docker within a Docker system thing anyway. But you know, like we don't at all touch the Data because we have a data

Speaker2: [01:17:27] Engineering team who

Speaker3: [01:17:29] You know has. He sets up the infrastructure like sets up the dags like, monitors the data quality like, we don't do anything with data. We're but we're essentially like one of the practical interfaces between the data scientist who operates in a dev environment and the services are doing engineering as well as at times, product. So, you know, it's interesting because I feel like this is one of the areas where, like everyone could kind of kind of come up with their own sort of like catch line or catch phrase, and it'll probably be wrong depending on the size, or it'll [01:18:00] be only loosely relevant depending on the size of the company and sort of the maturity of that company. Because if you're if you're a company that's not very mature small startup, then your big focus would be probably more on the data engineering side. And you would just build sort of like simple models that you could just even run the predictions and have them like surfaced and some kind of batch like,

Speaker2: [01:18:24] You know, pre cache sort of deal.

Speaker3: [01:18:27] Whereas in a bigger company, you are sort of concerned because you're treating all these ML products and your people as kind of assets that do need to be managed.

Speaker2: [01:18:35] So I don't know. I mean, that would be

Speaker3: [01:18:36] My sort of line would be like, it's just the discipline of delivering like ML products and doing them well. But but for some companies, like a startup, the ML versus the Data product, there might not actually be this

Speaker2: [01:18:50] Very clean sort of

Speaker3: [01:18:52] Line between what's a Data product versus what's a normal product versus what's just even having the product to begin with.

Harpreet: [01:19:00] Thank you very much. Let's go to Greg then after Greg, we're going to move into Gina's question, there's a question from coast. Then Lorraine on LinkedIn and then we'll go ahead and we'll call it a wrap after that. I go for it.

Speaker2: [01:19:14] I just have the probably the

Speaker4: [01:19:16] Most known technical answer here. But Joe was actually speaking my language.

Speaker2: [01:19:21] If you go deep into the history of of DevOps, it

Speaker4: [01:19:24] It it's built on top of lean, lean concepts. The Toyota Way is definitely something that will open up your eyes. To this, you will learn the concept of just in time. You'll learn the concept of reducing

Speaker2: [01:19:39] Variations down the production

Speaker4: [01:19:41] Line. So I see em a lot as a framework. You're in the manufacturing side, you have production lines, you're trying to reduce variability, you're trying to ensure reliability, you're trying to ensure efficiency.

Speaker2: [01:19:56] Therefore, you put this

Speaker4: [01:19:57] Framework called ML up so you can keep up with the demand. [01:20:00] So you think about your ML models, those are the supplies that you're pushing out to production to meet the demand of the business use cases. So with that, you have to find a way to constantly deliver them to these business needs so

Speaker2: [01:20:15] They can be addressed

Speaker4: [01:20:16] The right way and the only way to find out whether you're delivering

Speaker2: [01:20:20] Reliability

Speaker4: [01:20:21] And with efficiencies to be able to measure that right. So that's where you have the monitoring concept drifts that are,

Speaker2: [01:20:29] Et cetera, so you can make

Speaker4: [01:20:31] Sure you're tapping

Speaker2: [01:20:32] Into that reliability, right?

Speaker4: [01:20:34] Because people will ask you questions that model that I've just purchased from you to meet my business demand. Is it reliable? Is it a good for my use case? Does it respond to my need? And for that, you have to have those kind of processes.

Speaker2: [01:20:49] So it's all

Speaker4: [01:20:50] About repeatability, reproduce reproducibility and things like that. To me, it's like coming into a car manufacturing site and doing things

Speaker2: [01:21:01] Consistently and where you're

Speaker4: [01:21:02] Reducing your time to deployment through these practices. And there's a lot to learn from that helps. But also there's even more to learn from the lean practices as well. This gets me excited because I'm an industrial engineer and I've started in the lean systems, and that was a great question. I appreciate that.

Harpreet: [01:21:21] Greg, thank you. Ken, how you feel about those responses feeling good about that?

Speaker5: [01:21:24] Yeah, those are great. Thank you, everyone.

Harpreet: [01:21:26] So I'm gonna drop a link right here to the Ml Harp community in the chat I joined that can, if you'd like. There's a lot of good activity there, a lot of great conversations. You know, I'm just learning a lot looking at the questions people are asking, Are you going

Speaker4: [01:21:38] To get Demetrius on your podcast?

Harpreet: [01:21:40] We were talking about getting some going to

Speaker2: [01:21:43] Get something going. Yeah, definitely.

Harpreet: [01:21:45] Yeah, he's cool, man. Demetrios, cool. Cool guy. Let's go to. Let's go to Gina's question of the day after Gina will coast up on Lorraine on LinkedIn. I'll read the question.

Speaker3: [01:21:58] You? Are you still here? Oh hi, [01:22:00] hi. Can you guys hear me OK?

Harpreet: [01:22:02] Yep. Loud and clear.

Speaker3: [01:22:04] Excellent. So, yeah, first time joining this group, I learned about Harp and this happy hour from Super Data Science podcast. And so that was a very cool. And then I live in Davis, California, which is really close to Sacramento, which is where her is from originally. So that was pretty sweet, and I really enjoyed that podcast and

Speaker2: [01:22:27] Just go,

Speaker3: [01:22:28] Says smart marketing to Davis as well was cool. We'll have to catch up and I'm going to I'm going to try to keep this short first time person. I want to give a little context. So this is a job search question. Please, everybody don't do nowt or,

Speaker2: [01:22:44] You know,

Speaker3: [01:22:47] Job search questions. It's so tough. So background I have I have three degrees,

Speaker2: [01:22:54] So I'm OK with degrees in biology,

Speaker3: [01:22:58] Environmental studies, MBA. I am mid-career and so doing a career pivot. I've always been into analytical work and as data science tools, as I became more and more aware of data science tools coming up and up as compute power has grown so quickly and storage has become cheap. I started following this some years ago and then decided to quit my job and actually do a bootcamp, and so I really liked that there were some, some issues

Speaker4: [01:23:31] Earlier on, let's just say.

Speaker2: [01:23:34] But on the

Speaker3: [01:23:35] Whole, it's been very, very good. I because I hadn't really taken any vacations in a long time I had. I realized after I started bootcamp, I was really burnt out. So I've kind of taken my time as I've gone along in this and I really want to. I want my next move to be meaningful.

Speaker2: [01:23:55] I want to, you know,

Speaker3: [01:23:56] Not just take the first thing that comes along. Having said that, as [01:24:00] a bootcamp

Speaker2: [01:24:00] Grad, you know,

Speaker3: [01:24:02] I basically got my boot camp projects and I want to do some other projects, but it's almost like I'm in this weird spot of I'm willing to do just about anything, but I don't necessarily want to do it just about anywhere. I don't know if that makes sense. And so, so there's a challenge there of there are lots of Data science jobs out there, obviously. So focus is important.

Speaker2: [01:24:27] I'm, you know,

Speaker3: [01:24:28] Struggling with that a little bit. And of course, there's a whole, you know, every PDF has a long wish list. And how do you know which ones are for real and which? I mean, you can kind of tell. Sometimes I can read between the lines sometimes and see which ones are like, OK, which are just kind of putting everything in the kitchen sink in there versus, OK, this is kind of legit like they really need somebody who has five years of NLP experience, you know, could be a challenge to find, but I'm sure I know those people are out there. So my question, my question one is just around, you know, how best to use my time. And yes, networking is important, and LinkedIn is a very useful tool, et cetera. The second piece kind of someone was mentioning soft skills earlier and in the chat, and that's really important. And I think I bring those to the table as well as years and years of experience. So some people have suggested to me, Oh, you could be like kind of a team manager type person. And that might be good because I've done a lot of analytical work and I've worked in consulting.

Speaker3: [01:25:39] I've worked in a range of different environments and that experience brings some context and also allows me to talk to many different people. So someone was also talking about that earlier being able to communicate with folks across different areas in the company. So I guess another big question, [01:26:00] though, is, I mean, if you can put on your resume, I'm an excellent writer or I'm a great communicator and like, talk is cheap and I've seen I mean it. Just as an aside, I consider myself a very good writer and just very persnickety about language and punctuation and grammar and all the rest. And I worked with a guy who's a terrible writer. And on his resume, excellent writer. It's like, Oh, come on, give me a break. So you can only communicate so much on that resume. And so I would love. I know this is kind of a. You know, kind of a little bit of a data dump about me and and you know, where I'm at my job search, but I would so love to hear folks comments on this and that advice. Thank you in advance.

Harpreet: [01:26:45] I got you got a good lineup of people here to help answer. We're going to start off with Nick Nick, who is a career coach, best selling author extraordinaire. Then after Nick, we are going to go to Nicole Janeway bills, then

Speaker2: [01:26:57] Vivian, then Mexico. Then Greg, then Marc.

Harpreet: [01:27:00] Then can you get a lot of good advice? So if you are not taking notes already, do not fear because the artist Data Science Podcast is all transcribed. It's all recorded. It will be released on Sunday, and you can download the chat right here in the in the in the chat box here. Also shout out to Kate Strategy, who has just joined Kate, my friend. Get to see your living room or whatever it is that you're showing us an actual living room. Kate, let's go to a yeah, let's start off with Nick, then Nicole Janeway and then Vivian and Makiko, then Greg, then

Speaker2: [01:27:34] Marc, then can.

Speaker4: [01:27:38] Just for the sake of brevity, I know there's a bunch of different things you told us about, I think a few ideas I'm going to latch on to. And you can always just DM me on LinkedIn or something. You just need more help but build portfolio projects. That's probably why one big advice, right? You had mentioned something about, oh, like you had some projects. They quickly looked at your GitHub. We can talk about [01:28:00] that, but I think that there's building more is never a bad thing. And especially you're saying maybe having some challenges finding a job, but you've worked in consulting, you've done a few other things. Is there a stepping stone job like a job that's like in that industry? That's more data driven. So honestly, you being a great writer, it's great. But like, honestly, you're paid to write code and that's that's the end of it. And I'm someone who struggles with that, too.

Speaker2: [01:28:27] And that's why so many

Speaker4: [01:28:28] Interests and know I'm not a good fit for industry, and it's all because of that. But anyways, see if there's a job that's sort of in your industry that could be easier to step in stone, step step into and use that as a stepping stone and then build more projects. I think that's never a bad thing. And then as people mention, like burnouts, real like take care of yourself. But like if your version of like goofing off and taking time for yourself is building projects you're interested in, that's not a very bad way of spending your time. That's like awesome. So that's my quick pitch. But feel free to DM me because I know there's a lot going on.

Harpreet: [01:29:02] Nick, thank you very much. Let's go to Nicole and after Nicole Vivian, Mexico, then Greg, then Marc, then Ken.

Speaker3: [01:29:09] Nicole Oh man. Yeah, that was great advice, but that's tough to follow. So I guess I'll just say like echo everything that Nick said and also add, you know, in terms of exciting projects to you. One thing that you might consider is checking out the local Data portal. And I think one of the benefits to that is if you're looking for a job in your area and this is kind of

Speaker2: [01:29:40] Changed since our lives are

Speaker3: [01:29:42] A lot more remote right

Speaker2: [01:29:43] Now.

Speaker3: [01:29:44] But yeah, like pretty much, you know, when you're having these conversations with people who you meet in real life, real throwback there. But you know, they'll be curious about the results of your model and how that impacts them. And they'll want to hear like, Oh [01:30:00] yeah, well, what about my neighborhood? What about the neighborhood that I grew up in? So that's like a successful one strategy that I have used that I have just found a lot of people get excited about. And then another benefit to that is there's a lot of organizations that are hyper local and are just like looking for contractors and they they might be willing to even compensate you. And you can pair your data science skills with your your writing and be able to leverage both. And even like, I mean, you're not going to make like a, you know, a winning like thing salary writing for like a local think tank

Speaker2: [01:30:42] Or a like a

Speaker3: [01:30:45] Policy council or something like that. But like those, those roles are out there and some of them are are complicated, which which is fun. So to look, look for opportunities like that because it those I mean, having projects on your resume and your GitHub valuable, but having projects like with friends online like, that's really cool. So if you can plug into an existing team that's looking for contractors like that could be great. You can also offer contracting services in other places, too. So just the thought and some potential things to like get your wheels turning in terms of how to expand and not just be job searching. And then one final thing I just wanted to share with you is like, you know, just because someone tells you you might be a good fit for a job like, don't settle like if that's not what you see yourself doing long term, like, yeah, there's a difference between saying, OK, I'm going to take a job

Speaker2: [01:31:42] For a defined

Speaker3: [01:31:43] Period of time because I know it'll be the next step versus like getting stuck in a role because someone's like, Oh, you have these skills. But if they're not energizing to you and they don't leverage your technical abilities that you're you're real hard skills that you are, you know, hard [01:32:00] earned, especially as a career pivot or like, I know how hard that is. So yeah, just don't sell. That's that would be my including it.

Harpreet: [01:32:10] Cold, thank you, thank you so, so much. Bye. Nicole, if you want to leave

Speaker2: [01:32:14] A link in the chat

Harpreet: [01:32:16] For the amazing study guys that you have created for the CD MP exam,

Speaker2: [01:32:21] That would be awesome.

Harpreet: [01:32:22] Go ahead and drop that link. Also, everybody watching on LinkedIn go ahead and smash that like button on LinkedIn. I see you guys know you guys have smashed like also Kate scratchiness in the building. If you guys have not yet already, you must go and register for the dedicated conference. Yours truly will be there presenting an know Greg is going to be presenting. I know Joe. I believe it's going to be presenting. A lot of us are going to be out there. I'll be talking about strategic aspects of Mel Ops. Kate, thank you once again for putting on another amazing, dedicated conference. I might just start doing my own conference next year to just, you know, we'll see what happens. Let's go to let's go to Vivian, the Makiko, then Greg and Mark. Then can you get like I said, you know, you got a lot of advice coming your way. Don't worry, this is all transcribed and it will be on on the podcast for you take.

Speaker2: [01:33:17] Vivienne, are you still here? Oh, yeah,

Speaker3: [01:33:19] Sorry, I just was letting you finish, so I just feel like there's so many things I could say, so I'll just try to stick to a few things. For one, I just want you to. It sounded like when you were talking that you're kind of like thinking about what is important to you, like when you were saying like, Oh, I'll take like any job, but maybe the company is like more important to me or like the culture or something like that.

Speaker2: [01:33:43] And like, I think that's

Speaker3: [01:33:44] A really good place to start to be like that. Like because when I was so, I also did a boot camp and I was a career transition and I was very much like got to a place as I was interviewing and stuff where I realized like, Well, I don't [01:34:00] super want to be like, the culture matters a lot to me more than job title, maybe necessarily even because, like, I didn't want to be treated like a robot. Like, I didn't want to just be treated like a coding robot or something like I wanted to be someone who got to actually use like some of my decision making skills, especially like as a career pivot and having some like other experiences, I

Speaker2: [01:34:21] Felt like it brought.

Speaker3: [01:34:22] Like, like, there are companies that value people who have a wealth of experience because they bring

Speaker2: [01:34:31] Unique perspectives to the table.

Speaker3: [01:34:33] And that's like something they want. They want people with varied experience and stuff so like, know that those companies exist and like search for those companies and like, it may be difficult to feel like, how do I find them, though? So like just pay attention to like I don't know the kind of wording that they use in the in the job description. Also, like when you're being interviewed, you know, pay attention to like the methods they use when interviewing you like, well, so I work at Facebook. I got a job at Facebook and like something that just

Speaker2: [01:35:09] Made me like, want that

Speaker3: [01:35:10] Job even more like I was like, Oh my God, I got to have this job was because after every interview, the interviews were always very much like thinking, just thinking through problems. It wasn't about like, like, sure, there was a little bit of like coding and stuff just to see if I wasn't faking it, basically. But like a lot of it was like very focused on like seeing how I think as a person and like, am I a good problem solver? Like, am I somebody that could like, bring fresh perspectives to the table? And that was like something that was like, Oh, this is like the ultimate. I want to work here. I want this job, you know, so like, pay attention to that sort of thing. Also, when you're applying to jobs, my job coach for the program that I did very much encouraged me to not get too obsessed with, like not meeting the requirements on the job postings.

Speaker2: [01:35:59] And like, [01:36:00] this is especially

Speaker3: [01:36:01] A thing that he

Speaker2: [01:36:01] Had to like talk with me

Speaker3: [01:36:03] About because like men, for instance, are much more likely to just apply for jobs if they even if they only meet like 40 percent of the requirements or something, whereas women tend to, like hold themselves back until they meet, at least, you know, 80 percent or something like that. So like, like, pretend you're a white man, like, ask yourself what a white man would do and then like, have that confidence, you know? And I guess that, you know, like, like, recognize ways where you can, like, get out of your own way and just apply. Because also like if you don't like something that my job coach talked to me about a lot is he was like, even if you don't like perfectly need it if they end up liking you. There's a lot of like he has seen. A lot of people were like, they don't actually meet what the like, what the job description is asking for. But they liked that person so much that they then just like, changed the role to fit that person because they ended up liking them so much. So like, I guess that that's also something that I would bring to your attention. But yeah, like you being a career pivot. Like when I started seeing myself like a career picture as a strength instead of like a weakness or something like, oh, if only I was as experienced as other people or something. Like when I saw it as my strength, that's when things like really turned for me, like, you have a wealth of knowledge that is unique. That means that you are unique individual. That can be your superpower, you know.

Speaker2: [01:37:29] Anyway, hope that helps.

Harpreet: [01:37:32] Excellent advice, thank you so much. We've been at that point about career perimeters and just having that kind of outside view, it is crucial we need more of that in Data science. If everybody was just a statistician or just the coder or just mathematician, like we wouldn't have the point of views that we need at the table. Highly recommend checking out this book by

Speaker2: [01:37:53] Epstein

Harpreet: [01:37:54] Called Range. I don't know if you've seen that or

Speaker2: [01:37:57] Not or washed it or

Harpreet: [01:37:58] Whatever. Listen to it. Watch [01:38:00] and talk about it. Check that book out. I think you'll really appreciate that. We're going to go to Mexico then Greg,

Speaker2: [01:38:06] Denmark and then

Harpreet: [01:38:08] Coast up and Lorraine. We're going to have to table your questions because I kind of get dinner for the family and put the baby to sleep. So we'll go Mickey, Greg, then Marc. Then we'll call it her up. I mean, sorry, Mickey. Greg, Marc and then Ken then called her up.

Speaker3: [01:38:21] Yeah, so I just have to make specific recommendations on the networking part. So for me personally, I found networking to be one hundred percent super

Speaker2: [01:38:31] Useless regardless of whether

Speaker3: [01:38:32] Or not it was virtual or in person, if I really if I was trying to network. Um, if it was a group or meet up that I thought was super interesting, anyway, that is kind of like having an authentic interest in like what's going on. Helps more than anything else. So that's one thing I would say is that like there, some people will do this thing where they will just like hit up every Joe Jane, John Heap like on LinkedIn. Send a message.

Speaker2: [01:39:05] I'm not. I don't believe you would

Speaker3: [01:39:07] Do that necessarily, but I would just say that sometimes all these, especially these like speed networking events that can be kind of useless. I would say, like a first degree, all kind of like, what are these four groups or areas or topics you'd be super interested in and kind of filter off that? The second sort of recommendation is that so there are a bunch of different groups that I have found for me to be personally very, very useful as both a woman

Speaker2: [01:39:34] Of color who is also LGBTQ

Speaker3: [01:39:38] And specifically those groups were tech ladies. It's a they have a Facebook group, but more importantly, they have a job board for women or non-binary individuals where you can directly connect with the hiring manager. And it's a curated role. They are all roles that specifically they are making it a point to try to recruit from a diverse [01:40:00] pipeline, which is really, really nice. And there are a lot of career changes in that group.

Speaker2: [01:40:05] A lot of stories I've

Speaker3: [01:40:06] Seen are people who, you know, not just for data science, machine learning, but they decide they want to webdav. They picked up a class like two or three years ago, and they found a lot of great relationships within the tech ladies group, so I would recommend checking them out. Another group that could be really great is women and data science. Um, that is a fantastic group, and more importantly, they will host hackathons every year, which are typically social cause based. I think last year was more COVID related for obvious reasons. No, actually those women is in for

Speaker2: [01:40:43] Mortality in emergency

Speaker3: [01:40:45] Clinics. Either way, you get paired up with random people. I did it. It was fun. I even got paired up with some researchers at Hopkins and Harvard Med for us to do a study where we brought our data science and machine learning and coding skills, and they brought their domain knowledge and statistical knowledge, and we combined it to try to do a study on social determinants of health. So that's a really, really great example of how you can get a collaboration of people with different skill sets and experiences to produce really valuable good work. So I would check out women in science to other groups. You could take a look at our ML

Speaker2: [01:41:21] For good from Delta Analytics.

Speaker3: [01:41:23] They often have opportunities where you can do projects that are email for a good base and even code for America, because you know, if you're going to be adding of portfolio projects, especially if you already have a background in certain areas like environmental policy, those are areas that could use a lot of great work. My dad, for example, who for years cannot run an Excel sheet to save his life, which means our accounting was terrible for the family.

Speaker2: [01:41:49] So he's he's partnering with

Speaker3: [01:41:50] People to do a study on whale tracking patterns for the Coast Guard Auxiliary because what they want to understand is all these super boats that are coming in and getting stuck in like. [01:42:00] The canals, are they also going to be harming whale migration

Speaker2: [01:42:06] Patterns, which

Speaker3: [01:42:08] Has an environmental impact? Once again, this

Speaker2: [01:42:09] Guy could not for the life of

Speaker3: [01:42:11] Running an Excel sheet, he would've done very poorly in David Wagner's glasses. You know, so if you're more technical than my father? Blossom, you know, there's a lot of opportunities to have those kinds of collaborations have that kind of impact. Yes, I would to get those groups, I would definitely make sure when you work, you're really filtering out what you're super interested in and really focus on sort of authentic relationships. Yeah. And in terms of the roles, I would, you know, echo people's points like, don't feel like you have to settle if you see a job description where it's like, Oh man, this would take like four or five people to do, that's probably a red flag, you know? And definitely, definitely like, look at the culture and see if they churn and burn. They're like a data science machine learning engineering team

Speaker2: [01:43:01] Because that is very

Speaker3: [01:43:02] Real, like in some companies and some startups, you know? So yeah, I'd say, good luck.

Harpreet: [01:43:11] Thank you very much. We're going to go to market and Cannes after that, but just real quick shout out to Colleen. Eric Russell, I know we haven't heard from you guys, but I really appreciate the guy's presence here and thank you for hanging out. J.r. What's going on?

Speaker2: [01:43:24] Good to see you as well. And then also

Harpreet: [01:43:26] To us and thank you for hanging out

Speaker2: [01:43:28] And chilling.

Harpreet: [01:43:29] So let's go to Mark and then Ken. And if anybody else wants to chime in on this, we could do that. I know Greg left, but we'll go to Mark and then Ken.

Speaker2: [01:43:39] There's been so much great feedback.

Speaker4: [01:43:41] I've been learning as well from all the great comments, so it's been awesome.

Speaker2: [01:43:45] I'm hopefully going to try

Speaker4: [01:43:46] To provide a different perspective, not repeat their stuff, but essentially I just want to focus on the job search itself. And the way I really equate the job search is I can think of it as like a sales funnel. [01:44:00] I actually don't apply to the job descriptions. I think I personally think it's a waste of time because you have like hundreds of thousands of people just applying and it's kind of like putting your you're screaming into the void and instead you should be more tactical and strategic about how you approach that. So I think I think of the sales, the sales funnel sales funnel has kind of like the item model as the awareness, interest, desire in action. And I think those are different steps within the funnel. So awareness

Speaker2: [01:44:28] Is the hiring recruiter

Speaker4: [01:44:30] And hiring manager aware of you and the services you can provide

Speaker2: [01:44:33] As a business of one, i.e.

Speaker4: [01:44:36] You're trying to be employed. Interest is like the screening phone interview for this desire is the actual full interview and then action is like you sign the deal. And so through that you have various levers within the process, like where are you leaking in the funnel? You know, maybe you can get the interview, but if you crush the screening interview as well. But when you go to like the coding interview and like the super days, it's not doing so well, right? And so you can use those steps that form, like where do you focus your time on through the sales, the sales to the hiring process, but essentially you're selling yourself right, you're still you're selling these services that you provide to them. But I really want to focus on the awareness component. That's where you first start off is essentially. For awareness, there's two strategies that I use. One of them for the last time, a job search, I instead applied to a single job. I just create content LinkedIn with the goal of helping one person to build awareness and hiring managers to reach out to me. And that ended up doing multiple interviews in my current job at Adejumo. But the other side you can do is kind of like an SDR or business development representative is essentially you do like cold emails and really targeting, like you said, you're amazing writers. So this play to your strength. And so essentially, it's like I try to understand who is my customer as in like, who am I trying to sell my services to through which are my Data [01:46:00] skills, right? And so I identify which companies do I want to sell to and which companies sell to, and which companies are the my target type person to, who want to buy my services research and try to really profile and understand that.

Speaker4: [01:46:15] And then, oh, marketing funnel. Yeah, I think it might be the marketing, funnel, sales, funnel, marketing funnel, whatever it is reading the comments. But essentially from there is I start researching. So if I'm focused on startups, so I go to Crunchbase. So like where the current Series A Series B companies are going for bigger companies, right? But then LinkedIn is really powerful. So now I'm like searching for recruiters, hiring managers, you know, I'll read their post or they recently talking about posts. Are they recently saying, like talking about the space? And I use that as the entry to actually send an email or message them say, like, Hey, I see you have a job posting, you know, and you basically sell yourself saying, like, why I'm the best person for this. And the thing you have to remember is like, Yeah, it seems maybe sleazy would be like networking asking for something, but you're helping them because hiring is so freaking hard. And so you need to spoon feed all the amazing points that they'll bring back to their team as to why, like, you're the best person possible, be hired and essentially as you're making their job

Speaker2: [01:47:20] Easier by telling them

Speaker4: [01:47:22] Direct and being very targeted, like what's their pain points? How does your services as a Data professional solve that? And then why? Like, they should continue a conversation and the end of the email? I don't ask, like, do you want to be with me? I say, I'm interested in me with you, discuss more about this opportunity. And from there, that's how I go about my job search and really being really intentional.

Speaker2: [01:47:45] And so instead of saying like a hundred different

Speaker4: [01:47:47] Job applications, I'm really focused on certain people and really speaking to their needs.

Harpreet: [01:47:55] Mark, with the advice, thank you so much, Mark, appreciate that that was said that right there. Let's [01:48:00] go to let's go to Ken and then that you know what goes up since you're still here, man, we'll get to your question. I haven't. My wife hasn't come in yell at me yet, so I can still play.

Speaker5: [01:48:13] Well, I'll try to make it relatively short in typical fashion, I will probably provide some slightly contrarian advice. I'm interested to see what what the group thinks, but the first thing is related to how you describe that you're a good writer and someone else who you know is a terrible writer. Also said that they're a good writer. And there's that dissonance there, right? Do you have a link to a blog or something you've written on your resume? Aside from the resume itself?

Speaker3: [01:48:44] Yeah, I think, yeah, I do

Speaker2: [01:48:45] Think I have that on

Speaker3: [01:48:47] My blog and I don't want to I don't want to overstate the writing thing. I'm thinking maybe I shouldn't have said it in the same context as soft skills. But in any case, there are certain things that right, if you say it on a resume or even a cover letter, I mean, you know, again, talk is cheap. So how do you demonstrate that or how does that come across? And I think it's just like having links to blogs. So that's a great idea. I think just communicating with people over LinkedIn. But again, yeah, so I didn't want to. I don't want to overstate that too much, but please continue to hear your advice.

Speaker5: [01:49:26] No, I just wanted to create that emphasis on kind of showing in that way, at least for me, the resume is the living document.

Speaker2: [01:49:33] Now, I don't

Speaker5: [01:49:34] Think I've ever looked gone on someone's resume unless it was just like, Oh my God, this is so bad. Throw it in the pile without looking at their at least their GitHub, or if they had a personal website, at least

Speaker2: [01:49:44] Clicking through that, that's something that

Speaker5: [01:49:46] If it's there, I'm probably going to click on it. And that at least gives me some insight into perhaps their creativity or their writing ability or some of these other things. The second thing that that I would like to touch on is a lot of [01:50:00] people talked about not settling and finding the right role. I think that there is tremendous value in that, and I wouldn't want to detract from that. But I also think that there

Speaker2: [01:50:09] Is a lot of

Speaker5: [01:50:10] Value in having a role within this domain to begin with, because that gives you so many more

Speaker2: [01:50:16] Opportunities to land

Speaker5: [01:50:17] Other jobs within the domain, like once someone has their first Data science job. It is so much easier to land a second Data science job in whatever domain they would want.

Speaker2: [01:50:26] So I would want to say that like, absolutely don't settle. But unless you have a

Speaker5: [01:50:32] Large, like crazy, large time frame or in theory, you're OK with being like a little bit inefficient with your time. I would say it's there's a premium on landing a job sooner rather than later, just because you can accumulate those skills and then you could transition into the larger pool of possible positions that are available to you after you have that initial position. You know, I would also say there isn't a stigma, at least right now, with changing roles in this domain. You can go in, you can create value for a company and in six months in a year in a pretty short period of time, you can go somewhere else without having any negative affect associated with you. If you did that five, six times or maybe even three or four times, that could be a problem. But from what I'm seeing, most companies do not have a problem with that in this day and age. And so

Speaker2: [01:51:26] The last

Speaker5: [01:51:26] Thing I would say is it is kind of difficult to tell if a company is

Speaker2: [01:51:31] A great fit without working there.

Speaker5: [01:51:34] Honestly, I've been in interviews where everything looked great on paper, and then once I actually got into the company, I was like, Oh, this is

Speaker2: [01:51:42] This is not as advertised.

Speaker5: [01:51:44] And so I think that there are opportunities there. But one last thing is that

Speaker2: [01:51:50] At any time, I think internships

Speaker5: [01:51:52] Are a possibility. I think Nicole mentioned contract work, if I recall as well. And those are great ways to get [01:52:00] at least introduced to a company to get familiar with the work, to see if it would be a good fit and also build experience. So hopefully that wasn't too much too fast. But those are kind of my again, slightly contrarian takes on on on your specific situation here. I hope it is helpful.

Harpreet: [01:52:18] And thank you so much. And like, if anybody ever comes to asking you why you've been switching jobs every six to nine months for the last two years,

Speaker2: [01:52:26] You just ask them

Harpreet: [01:52:27] If they've ever been in a position to turn down opportunities as they come to them.

Speaker2: [01:52:32] Yes, I sure as hell have not been in

Harpreet: [01:52:36] Positions where I just turned down opportunities and they come to me. People keep coming to me. The job offers, what do you want me to do? They know, and then just leave it at that cost this go to your question.

Speaker4: [01:52:49] Sorry, I'm on mute.

Speaker2: [01:52:50] Just before that, Jane. Just one thing that

Speaker4: [01:52:54] May be worth thinking about is exactly how you structure that resume now. If you're applying to a small to midsize company where a hiring manager has time to see seven or eight resumes. You would structure your resume a little bit differently, but

Speaker2: [01:53:09] If there's 100 people

Speaker4: [01:53:11] Applying for the same role, right? And right now, there's so many people coming out of graduate school and things like that with data science and interest and data science roles, and that analytics roles, you've got to make it stand out. Like I designed this thing like a flier, right? Like think of a movie flier.

Speaker2: [01:53:25] This is the you're the you're the Showtime.

Speaker4: [01:53:28] You're the real show, right? You've got to make it super easy, like when you see the next Avengers post to come out, you know, when it's

Speaker2: [01:53:35] Coming out, you know

Speaker4: [01:53:36] Which character it's about.

Speaker2: [01:53:38] You know, all of these things,

Speaker4: [01:53:39] You can glance at it and you'll know that information, right? So I kind of keep like a seven second rule with my resumes where I'm kind of like it should. If a hiring manager accidentally sees my resume on their desk or on their colleagues desk, they should gather enough information from it that they want to read the rest of it, that they want to go see my GitHub, that they want to go see my website [01:54:00] and things like that. And then that builds that conversation, right? So if that's the early block that you're having trouble landing those interviews with,

Speaker2: [01:54:07] Yeah, I'd look

Speaker4: [01:54:08] At how you're that early part of that funnel that Mike was working on. How do you attract attention to it? Right. There's so much. And to be fair, I don't know if it's good advice or bad advice because

Speaker2: [01:54:18] My resume could end up

Speaker4: [01:54:19] Looking pretty

Speaker2: [01:54:20] Zany and they have

Speaker4: [01:54:21] In the past. But I think there's value

Speaker2: [01:54:24] In it and it's different. So you definitely

Speaker4: [01:54:26] Stand out for the veteran, who is really my question. I think we've touched on some aspects of it, but the rest of it is could actually be a pretty deep rabbit hole.

Speaker2: [01:54:37] But I guess help line in as you're trying to

Speaker4: [01:54:40] Grow a ml engineering for a product and a team. What's like what are some real like bear traps that you've seen people you guys have seen people walk straight into like common

Speaker2: [01:54:53] Bear traps are

Speaker4: [01:54:54] Like I said, this could be I'm talking about a mid scale company. Let's let's, let's scope, define this to you. Finished your pilot and you're looking to scale into a product, right? What are the things that people miss?

Harpreet: [01:55:08] So just. Rephrase that question, saying if people are scaling their team,

Speaker4: [01:55:14] So I mean, the reason I'm stopping this question is because I mean, it's a different answer to early stage research start up to like a large scale company that's already running, right? But let's say you've finished your pilot and you've done it with maybe one or two data scientists and you'll look to your first major launch, right? But you need to build out your team for that. What's what are some of the

Speaker2: [01:55:36] Big,

Speaker4: [01:55:38] Big mistakes or preconceived notions that people go into it thinking, Oh, I need to build this big and I need to scale. So I obviously need X, Y and Z, but they actually don't flash. Oh, I can't possibly need to do a B and C, but in reality, it's in depth. Extremely important. Like I said, this could be a big rabbit hole, so it might be worth saving this [01:56:00] for another podcast.

Harpreet: [01:56:01] Yeah, definitely. Anybody wants to chime in here. Let me know. I'm trying to.

Speaker2: [01:56:03] I'm trying to like

Harpreet: [01:56:05] There's only the question here.

Speaker2: [01:56:06] Real quick, though the question

Harpreet: [01:56:07] Does it have to do with OK, like, you know, you're a fledgling team with some number of people getting some experience under your belt and you're starting to over scope too quickly thinking that you need to do this, that and the other thing, but you actually don't.

Speaker2: [01:56:23] Or is it OK,

Harpreet: [01:56:25] You're a fledgling

Speaker2: [01:56:26] Team who's beginning to

Harpreet: [01:56:28] Grow larger and larger? Here are some of the challenges you might run into. You're not

Speaker2: [01:56:32] Careful which way you want

Harpreet: [01:56:34] To take that question.

Speaker2: [01:56:35] Probably the probably

Speaker4: [01:56:38] The latter

Speaker2: [01:56:40] Where there is actual potential

Speaker4: [01:56:41] For scope. It's just the team, as it exists, is still trying to build that out.

Harpreet: [01:56:48] Yeah. I have a few tips, I guess, and then if anybody else wants to chime in, go ahead and just raise your hand there. But I mean.

Speaker2: [01:56:56] Inefficient workflows, I think,

Harpreet: [01:56:58] Is one thing, like when you got manual processes every step of the way, from data extraction to cleaning to modeling to deployment, everything is manually and done with custom-made lines of code. Or, you know, if you're still copying code between projects you don't have, like a central repo of useful code snippets and you have to like reinvent the wheel for every project. That's another pitfall. I mean, sometimes you

Speaker2: [01:57:24] Just end up with just really long

Harpreet: [01:57:26] Ramp up times,

Speaker2: [01:57:28] You know, get get started, you know,

Harpreet: [01:57:30] Cooling and then stuff like that, if you have if you're relying overly on software, expertize and other teams. I mean, there's other stuff that you should worry about long, long retraining cycles and things like that because, yeah, go for it. Clarify.

Speaker4: [01:57:50] I mean, so it's to a point. It's like I've seen situations where there's this, there's potentially a preconceived notion that, Hey, we need this, [01:58:00] we need this fully automated training experimentation pipeline when in reality, it's the Data extraction that needs automation, for example. Right? Actually, running the experiments, you might not need to run that many experiments because you've already kind of got a model that did the proof of concept

Speaker2: [01:58:19] And you're going to improve that first.

Speaker4: [01:58:21] You now need this Data pipeline that's coming through, right? Like so that's potentially one situation where there might be too much of an emphasis on the experimentation being automated. And, you know, like, do you see that? Is it a pattern that you've noticed that people continually misidentify the bits that need to be

Speaker2: [01:58:43] Optimized in a

Speaker4: [01:58:45] In a project or in a team? And where are

Speaker2: [01:58:48] Those?

Speaker4: [01:58:49] And one of those things that we usually misidentified.

Harpreet: [01:58:53] Me, I'd love to hear from.

Speaker2: [01:58:55] I think Mark

Harpreet: [01:58:55] Or Nick might have some insight here, hopefully. I mean, the whole thing about collaboration can get hard right, especially like, you know, machine learning is more than just code Data. Science is more than just code. It's not like software engineering where we just, you know, commit code to get or whatever. We still got to

Speaker2: [01:59:13] Worry about, like you mentioned,

Harpreet: [01:59:15] Pipelines and experiments and Harp

Speaker2: [01:59:16] Parameters and all that stuff. So just

Harpreet: [01:59:18] Facilitating collaboration. If you could find ways to make that easier from the jump, that will be crucial and an extremely helpful. I mean. Let's just leave. Oh, you. Other people are leaving. But yeah,

Speaker2: [01:59:36] Let's see if like if Mark

Harpreet: [01:59:37] Got any insights on that or any tips.

Speaker4: [01:59:41] Yeah, look, to be honest, to be honest, I think I think the question is a little bit overly broad right now, purely like for the lack of time. This might be something

Speaker2: [01:59:51] That is better. Like, defined

Speaker4: [01:59:53] It. They'll discuss broader.

Speaker2: [01:59:55] Yeah, I mean, yeah, if you got any

Harpreet: [01:59:57] Insights, let me know. If not, then it's [02:00:00] all good.

Speaker4: [02:00:00] We can all wrap it up. I was just about to ask two clarifying questions. I can give better advice. I already come up with about 10 questions that I could ask myself to clarify this, that's not the right time. But look, I think the key thing I got from there is like, at least from a starter perspective, I focus on prioritization.

Speaker2: [02:00:20] That's the thing

Speaker4: [02:00:21] I had to get really good at. And so there's this tradeoff between am I going to build essentially frameworks and documentation to make this process smoother and repeatable or go on to more impact

Speaker2: [02:00:34] Built a new feature

Speaker4: [02:00:35] That will get us more money. And there's essentially if I could put in there for a second, essentially, what you're saying is you need someone with a real understanding of what the product needs are right? Like, you almost need someone who's got those product management skills and be able to identify what's actually necessary. Yeah. So for for me, there's always a cost going back and forth of like, all right, I I built this tool and now customers are using it, and now they're requesting all these custom analytics. And now I'm doing custom analytics and the less feature building. And because I'm doing these custom analytics all the time now, I have to build out documentation and tooling to repeat this process again. Then it became a priority for me to be like, Oh actually, Larry, let me just go create how this repo where you create these functions that we can all the other data scientists can pull from as well. And so, yeah, I think it really goes back down to like where where's the real big priority for you? And like, I really learned this. I had a whole conversation. I think months ago. Talking about prioritization is, you know, Watson, get my boss a promotion wasn't get me a promotion and I can say no to a lot of things and only focus on those high impact things. And once I start doing that, my life became a lot easier because like now, I'm doing less work, but now I'm driving way more impact. And many times those those kind of like structural things aren't aren't the thing that gets me the promotion, but [02:02:00] some things are. So it really depends, and it's really being critical. Like, where are the top three things for this quarter that's going to really drive that? So like last quarter, I was like building data access. So that was really a lot of internal tooling and making this process repeatable and dependable. But other other quarters, it was

Speaker2: [02:02:18] Literally like, you

Speaker4: [02:02:19] Know, put out this new feature for this customer or for our customers in our product, right? And I don't care about documentation at that point that I'm just focused on getting this feature out.

Harpreet: [02:02:30] Mark, thank you. Close up, yeah, great question, like definitely type this out. Send it to me. I would love to get this going on like a LinkedIn thread. I think a lot of us could do that value could tackle the all the OGS Jovan, you know, as well. Dave Langer, speaking of I haven't seen Dave Langer like six months, man.

Speaker2: [02:02:47] Dave, miss you. Come back

Harpreet: [02:02:49] Then. Been a while as well. Let's go ahead and wrap this up. Thank you guys so much for hanging out 52 weeks. That's one year we've been out here. Doing this could not have done it without you guys. It probably would not have been as good as it has

Speaker2: [02:03:02] Been without you

Harpreet: [02:03:03] Guys. I'm grateful, eternally grateful to every single one of you, everyone who's been coming. I know there's a lot of people that that that I haven't seen in a while, and I miss you guys. Hopefully, you make time during the holiday season to come and hang out.

Speaker2: [02:03:17] You announcements,

Harpreet: [02:03:18] Though do not forget that I'm going live on LinkedIn three times in the next seven days. So Brant dike's tomorrow we'll be talking about Data storytelling that is at 11:00 a.m. Central Time with Brant Dikes on Wednesday. At some point during the afternoon central time on Wednesday, I'll be talking to Joe

Speaker2: [02:03:37] Reese live

Harpreet: [02:03:38] On on on LinkedIn as well. And then the following Saturday. That's October 9th. Talking to Brittney, though, we're going to be talking about Brittany's book

Speaker2: [02:03:48] Bigger than leadership.

Harpreet: [02:03:50] So looking forward to talking to Brittany about that. Also, do not forget to sign up for Data cated conference. Kate Strachan. Thank you for hanging out for a few minutes. Definitely [02:04:00] go check out that conference. I'll be

Speaker2: [02:04:02] Presenting a team presenting

Harpreet: [02:04:04] A presentation about how you can strategize matchups for your teams, so not necessarily get into technical details.

Speaker2: [02:04:12] But how about when it's a good time to

Harpreet: [02:04:16] Start looking at and MLPs strategy?

Speaker2: [02:04:19] Also, I was on

Harpreet: [02:04:21] The Narrative Science podcast. Hopefully you guys get a chance to check that out. I was on there with Kate Cassidy Shields, which is huge for me. Narrative sciences, that's a huge company.

Speaker2: [02:04:32] And just to get them to

Harpreet: [02:04:33] Want to talk to me, whatever reason they thought, it would be a good idea to have me on.

Speaker2: [02:04:38] That was cool. I really appreciated that.

Harpreet: [02:04:40] Thank you, Cassidy, for having me on. Also, I booked an interview with. The author of the Creativity Code, Marcus du Fatwa, he is the public facing mathematician at Oxford University. He's all over the BBC doing all these documentaries and things like that, so I'm excited about that. We'll be talking mostly about his book Creativity.

Speaker2: [02:05:03] This book is

Harpreet: [02:05:04] That creativity code it's all about. A deep learning's ability to help augment human creativity. And also, we'll be talking about his new book that is coming out as well as I'm excited for that. Do you think he has so much one year? Let's keep it going for the next 52 weeks? Thank you guys so much for being here. Thank you guys. So much for hanging out. Another huge episode. My friends remember you've got one life on this planet. Why not try to do something big? Here's everyone.