The Artists of Data Science - Data Science Happy Hour #72.mp3

Harpreet: [00:00:09] Let's go. What's up, everybody, welcome to the artist and Data science. Happy hour. It is Friday, March 11th. I'm your host for the next few minutes. Harpreet Sahota before I hand it off to the one and only community member Mark Freedman. Super excited to have Marc take over the reins today. Just a quick shout out for you all.

Speaker2: [00:00:29] Released an episode today

Speaker3: [00:00:30] On the podcast with Brant Dikes. This is all about storytelling, Data storytelling, so please do check out that episode. I think I think he has learn a lot from it. We had a lot of great episodes drop like this month in general last weekend for Method or just talking about how he broke into

Speaker2: [00:00:48] Into Data science did one with with Justin.

Harpreet: [00:00:52] When we're talking about, well, just the ones the guy that does the Declassified

Speaker3: [00:00:57] College podcast did a couple of other podcasts episode with Liz Foxley, coauthor of No Hard Feelings and

Harpreet: [00:01:05] Another One with Alyssa

Speaker2: [00:01:07] Simpson Ross. So definitely lots of great episodes to check

Speaker3: [00:01:10] Out on the podcast. If you have not yet checked them out, please do. But the newest one is with Brant Dikes.

Speaker2: [00:01:16] Also, huge shout out to everyone who supported

Speaker3: [00:01:20] With the Buy Me a coffee initiative that I set up to kind of help help kind of get stuff back up and running.

Speaker2: [00:01:28] Shout out to Matt Dymond, Russell Willis, Russell Lewis with the massive support.

Speaker3: [00:01:33] Russell, that was the incredible man. Thank you so much for that love. Schwab also cooked it up as well as you guys. Thank you so much for your support and kind of help me get back up and running while we figure out this stuff with the basement and the insurance and all that stuff you guys support is much appreciated.

Speaker2: [00:01:52] So thank you so much. If you guys want to help

Speaker3: [00:01:54] Out, there is always a link, right in every podcast episode to buy

Speaker2: [00:01:59] Me [00:02:00] a coffee

Speaker3: [00:02:00] Bean. Go ahead, check it out whatever you can to help me to use that stuff, to buy me a new microphone first and foremost, and some headphones so that the quality is good.

Speaker2: [00:02:11] Other than that man, I must sit back

Speaker3: [00:02:13] For just a couple of minutes because I miss all the awesome and the chill, but I'm having the reins over to mark who will be your host for the rest of this evening. Mark, thank you so much, man, for taking over. I appreciate you stepping up and help them out, man.

Speaker2: [00:02:28] Of course. What's good, everyone? Mark Freeman, you've shown up here. You probably see me quite a bit because I come here almost weekly. But, you know, opportunity to help out Harp any time that pops up, I love to. He's done so much for my career, this space, so always happy to be here and support whenever I can also hang out with these lovely Data folks and Data nerds and also seeing people on the LinkedIn. Shout out Sarah Maram celebrating that you've got a second interview. How amazing. So I will be keeping the LinkedIn if you have questions over there and we have questions here, you know, happy to have that conversation. And so to kind of kick us off first. One of my first questions, I put out a post today on LinkedIn asking about what's your Data dream team? You know, many times there's a lot of Data initiatives where companies like We need Data, we need to have all these initiatives and then they hire someone. They kind of goes wrong. So you know what, if you had kind of control, you know what, Data dream team you would want to have as maybe as a leader or if you're someone new and trying to think, you know what type of person I want to be on a new team, you know, what would that dream team be for you? For me personally, I said first would be Data analyst, then Data engineer and then data scientist, finally. But there was some debate in the comments, so I'm curious what others have to say and their perspectives.

Speaker2: [00:03:58] And so to kind of kick it off, I'm [00:04:00] just going to go with the easy ask. Then, you know, where are your thoughts, you know, since you've built teams before you start off? Mark Cuban leading the team and funding it. So it's possible Mark Cuban, if I could get the pearl to hit everything up as far as the research side of things, I would go with that definitely entering. You've got to have him. He's the one who's going to be doing, you know, data literacy and education throughout the organization. So I'm bringing him in. But you can hear like I'm calling out these roles that nobody talks about. You know, I'm talking about somebody who's going to be saying yes to innovation, because if you don't have that person, you're done. I mean, just pack it up while I have a team. If you don't have someone who's going to run the educational side of things, who's going to not only teach the team and grow and mentor the team, but also the rest of the organization, then you're in trouble there, too. And if you don't have someone who knows how to run research somebody that has an academic background, but it's also like made money with research, again, you're set up to fail. So when we talk about all of these roles within the team, we always forget the ones outside. And there's all these other pieces that need to be in place or your data scientists just kind of shows up and goes, where's the data? And even if there's a ton of it, it doesn't matter.

Speaker2: [00:05:18] You know, your data engineer is like, All right, I'm going to start gathering data. Why would we gathering data for? And so, you know, you need your data engineer. You need your data scientists, you need a machine learning engineer, somebody ml ops, you need a product manager, but then you also need all of these connections to the business. You need to figure out a way to scale, a way to grow. And the more I think about it, you know, trying to bring in one or two people to kick things off. I don't know if that can work anymore. When I first started, it was, you know, solo I was the data science team. But like in 2014, data science was really easy. It was kind of over glorified analytics, and big data really [00:06:00] wasn't as big as it is now. And the field isn't as broad. So I'm wondering, I just don't think you can do it with one or two people. I think you need to. I mean, you need the analytics foundation. You need the Data Foundation first. So if you don't have that, obviously get there, but you almost immediately from there, take the full plunge and go off the cliff. Not, you know, down the stairs. So I would actually want to follow up on that on that kind of statement, is you saying it's requiring a team that requires a huge investments? Oh, you know, to to have that. So, you know, people can really kind of say like, Hey, you know, we can hire a couple of people for head count, but to have a full department.

Speaker2: [00:06:43] That's a huge ask how how does someone kind of make that buy in for leadership to actually assess the budget even do so because that's a huge risk. That's Mark Cuban. That's that was that first person that I said to bring in. Because if the team's not working towards something innovative, if the team's not working towards building things that will generate significant revenue because free cash flow right now is hard to find, companies are struggling for free cash flow and they're getting punished. If you do like long term guidance where you say, Well, you know, I'm going to have to guide down a little bit, your stock price goes down 20 percent. It's wild. I mean, tech right now, valuations are getting pummeled. And so this is the dilemma. If you can't have someone who has that long term vision who's willing to, you know, Satya Nadella is another great example of this of someone who can come in and explain, Look, we're all in, and here's why, because this is the revenue stream of the future. Here's why we're putting all this cash in and doing it the right way. Because if you look at the way Microsoft tried to get into cloud, you know they did it sideways and they didn't put a whole lot of effort into it. In Azure, it was kind of a train wreck to begin with, and then you had to take over. And he basically said that [00:08:00] like, I'm all in on cloud.

Speaker2: [00:08:02] And so as a company, even as a small company, if you're getting into data science, you have to have someone who has the vision who understands not only how the product is going to work, but really how the business has to transform in order to support it. And you have to have this concept of revenue like crazy amounts of revenue, not just over the next six months, but looking out five, six years. If you don't have that sort of a vision, if you don't have that Mark Cuban type where the Satya Nadella type, you're in a lot of trouble because that investment, you're totally right. There's no way to justify all of and we haven't even talked about like the infrastructure money we've been talking about, all of that transformation that costs money and you have to bring in leadership and expertize. There's a lot to be done if you don't have that revenue stream, you know, pot of gold at the end of the rainbow, there's just no way for the business to justify everything you have to do. You end up going half way and you lose more money because you're spending. It's still, you know, two or three people and the infrastructure over the course of a couple of years. It's a lot of money. And if you don't get returns, then you're just sinking three years of time or two years of time and all that cash. Awesome. I just want to quickly ask other people questions if anyone else has thoughts on this question of this Data dream team.

Speaker3: [00:09:23] I think sometimes people overlook that sometimes the talent already is there. They're just not. They're not given the mandate or the space to do the good work that needs to be done. But we have kind of experience on my team where we went through this massive change in the company and before that, a lot of work was definitely valuable. I'm not going to say the work is never valuable, but it did feel sometimes like we were spinning our wheels on, like fighting mini fires as opposed to building like more of [00:10:00] a fireproof house. And there is there is a big shakeout. We lost a lot of people. It forced the company or at least the company and also the board to really kind of rethink how we were spending time. So the first thing that happened was we cut out unnecessary projects. Then we said, OK, we're going to really clarify the work in the direction of our team and we're going to and by we I mean, like my my manager was a big part of this and agitating to our director saying like, we have good people here, we also want to keep them. But more importantly, being good service people is not enabling us to actually do work. That's impactful. Like in some ways, being the golden retriever is just like it wasn't serving the business or Data scientists. And so she did a lot of work and advocating both for the executives above, but also getting input from us to serve what we wanted as a team.

Speaker3: [00:11:02] And so sometimes there's a lot of like unlock talent that's already there, but you still need like somehow like there needs to be some kind of change. And unfortunately, a lot of times that transformational change either comes from like bringing someone external in or it comes from like a crisis point. You know, where then? Oh, yeah, like, you know, we need to do it, and sometimes it's like people know they should have been doing things a better way, but they they just need an excuse like, oh, like we lost half our people. So now we can't do things the same because it's like, well, that work was always kind of low value. So if you could just kind of get tossed out the window, then why were we ever doing that work? But so I think that's something to kind of think about too, like just to just to play devil's advocate. But, you know, like, I would love to have a Mark Cuban, right? I mean, who else is who? Who else is going to rep better than Mark Cuban? Oh, [00:12:00] before I bounce out and and give mark full range, I'm wondering how much the maturity of the Data maturity of the organization would impact the starting lineup of that Data team, right?

Speaker2: [00:12:14] So I've been like the founding member of Data team like twice and

Speaker3: [00:12:17] Both times zero Data maturity. And it's been the both times extremely painful. And, you know, I'm wondering how that changes some people's responses, but I think either way, I think one of the first hires, if not the first one, should be like a Data savvy product manager who's already maybe

Speaker2: [00:12:39] Familiar with the company or who's already internal to the company who can really advocate for some of the work that the other people might

Speaker3: [00:12:47] Need to do. And then definitely, I think without a doubt, the second hiring seems like a Data engineer because Data scientists, Data analysts were end users of Data or like an analytics engineer,

Speaker2: [00:12:59] One of those hybrids.

Speaker3: [00:13:00] Because without that infrastructure in place, we can't get work done efficiently or effectively. I'm going to catch up on the discussions in and here I go, but I'm interested to see see some

Harpreet: [00:13:16] Responses to that.

Speaker2: [00:13:17] And I totally agree with VIN statement to clone Greg. If we can all clone great, we'd all be set Data careers. And also just a quick reminder to if you have any questions, please message me in the chat or put in Link LinkedIn and I'll put you in the queue. And so we'll go with Russell first for your response.

Speaker4: [00:13:38] Thanks, mom. Yes, I think great input already. I think Ben's comments were on the mark felt organizations that are large have a big war chest or budget and have a lot of bandwidth of workforce that they can allocate to the problem. But say, for a very young organization that's, you know, in bed in [00:14:00] the double figures for the organization, I think Harpreet Sahota comments are more appropriate for that size and whether it be a Data savvy product manager or a business savvy Data person. I think there's there's a threshold there that, you know, with a with a good crossover of, say, middle 50 percent that could go either way. But yeah, the scale of the organization is a significant criteria in making that decision. And as far as Data maturity and the organization's go, I think that's also a big thing. You know, if especially if it's a large organization, if there has been ingrained paradigms and habits of doing things a certain way, trying to break those and take them down a structured, data driven Data enlightened approach, there can be an awful lot of inertia to break that direction and move you out of that rut, and that can be quite a painful process. But as soon as you start to move out, as soon as you reach direct that inertia, it gets easier. Every every increment that you take towards a new direction gets easier. So there's a lot of what it sounds like a simple question. There's actually a lot of permutations to that. But I like your approach with the with the analyst versus the engineer for the for the the low scale early stuff. But, you know, for a big operation, perhaps the engineer first. And it also depends how much data is available at the outset. So not just the quality, but the the quantity of data available also.

Speaker2: [00:15:45] So we'll go over to you because up.

Harpreet: [00:15:49] So I mean, this goes back to what Harpreet was saying and kind of also throws back to a couple of the videos then that you posted on your YouTube channel [00:16:00] maybe a year ago, right? How do you get buy in from a company? Right. That's the biggest thing you need buy-in from from leadership. If they're not seeing the value in it, they're not investing the money in it. If they're not investing the money in it, you're not going to get a team up and running to do some serious work, right? It doesn't matter how good you are. There is a physical and real limit to how much impact you have on a huge organization from a Data perspective as an individual. Eventually, you're going to need a team around you, including Data engineers, including infrastructure that can help you sort everything out that you need to host models and train models to sort out your Data. So how do you get that buy in? Do the people at the top actually understand the value of the data that they're sitting on? Do they understand the cleanliness of the data that they're sitting on the potential business impact that it can have if they don't understand that? No, you're not going to be able to convince them to put together a full team. Right. And that's where it takes, you know, experience data scientists. And maybe this is this is what I started believing in consulting and machine learning and data science as a consulting thing.

Harpreet: [00:17:11] So many times a company might say, Oh yes, we need machine learning, or yes, we need data science, right? Let's do it. But then in order to do it internally, you maybe have one person who's there in the company as a VP for like, you know, six months before they move on to their next big career move. And they want to say that they let a data science initiative. So they hired two data scientists, put them in the ship, give them pretty much nothing and then watch it flail. It survives long enough for them to put it on a resume and move on, but it doesn't have any real impact, right? And you see it a billion times. I'm sorry language, but geez, you see it all the time, right? Cut to the chase. That's when you need a consultant to come in and give you that significant expertize at a management level, at a C-suite level to say, Hey, this is the data you're sitting on. Here's the value [00:18:00] build the business case. If you can build the business case, I can guarantee pretty much any C-suite or executive team is going to sit up and take notice until you can build the business case. They don't give a damn and fair enough to them. Right? So yeah, what do you need to build the business case to give you that full team? That's how you get buy-in, right?

Speaker2: [00:18:20] And I'm new to this hosting stuff, so I'm trying to figure out how to get back into regular view because I'll put you on spotlight and now it's not coming off. So we all learning together here. It's great. All right, we're back now. And yeah, I agree. I think in my my original. I initially said I think the Data analysts should be the first hire, but after hearing Vince kind of talk about how it should be not individual hires or a whole team, that's that definitely changes my perspective. But I would caveat is that I still think that Data analysts should be a first hire and not for a full initiative for for data science, but as someone to start building that use case in the Data to have that report to bring to leadership. And so, you know, I'll go ahead.

Harpreet: [00:19:07] Could I ask, is that what you consider a first hire or is that what you'd go to a consultant for? Because you could achieve the same thing with a consultant, right? Which gives you that short term ability to experiment to figure out, Hey, what's the potential value in here? Right? And then you figure out, OK, now we're going to create a data science team internally, then you already know the value, you know the gap. So the consultant can give you an analysis of, Hey, here's your infrastructure gap. Here's your engineering setup gap, right? Then you start going, OK, now I need the engineers in place to build the foundational platform that can let our model builders or Data sciences come in and actually their analysts come in and make the real value right. Because too many times you see companies that have hired a data analyst or a data scientist, first, they build great models on poor, foundational data structures, poor foundational, [00:20:00] you know, data engineering flows, workflows, ETL, whatever you want to call it, right? Too many times you see that foundational bit just missing. And then it takes a couple of years of that and then it's just not really giving the real time benefits to say, OK, crap, we need a full data engineering team to flesh out that. And then you go through the situation where you've got these janky models that have been deployed for like a year, and then the whole company has got to take this big initiative to refactor the whole damn thing, right?

Speaker2: [00:20:31] So it's funny that you bring that up because one of the one the commenters on my post said exactly that, and they had the background to back it up as well, to say, like, I've jumped into multiple teams where the data analyst was hired first and they created this a lot of technical debt. And then they had to she had to come in and fix everything, right? And I actually I agree with that, and I just want to pause it like, you know, potentially I hear the hear the argument of like, why hire Data analysts when you can hire consultants? And I think the big thing too is like this buying component. And so again, I think I should probably reference to other consultants or people who have hired consultants. When a consultant comes in, how hard is it for when they give their recommendation for the whole org to accept it and want to move forward with it? You know, I think there are some wins with that, but I feel like someone who's within the org and who's working with those. Those people building that rapport over a long term can start getting to those real questions that a consultant jumping in and jumping out may necessarily not be able to do.

Speaker2: [00:21:37] I think consultants are amazing. I think that, you know, for the right case, who has some consultants here. But you know, that's that's the argument I'm making is that there's this long term rapport and they're starting to build up this case. And the Data ANL is more so. Instead of trying to build its infrastructure, they're more tactical of like, where are the high or the low hanging fruit to make people [00:22:00] start caring? I think the Data analyst is, rather than being an actual infrastructure and huge piece, they're more a catalyst to actually get the Data, the company to be like, Oh, actually, this Data thing is actually really cool. We should start paying attention and oh, I want more data, but we all have all these pain points, right? Then I think it would be worthwhile to kind of bring in the consultant, but I'm open to any like pushback on this as well. And I think you have some thoughts.

Harpreet: [00:22:25] One hundred percent. I've got some thoughts on this one. All right. So two questions. This sounds like a different problem, right? So if you're hiring a consultant and you're not willing to take their feedback seriously, eh? Have you hired a consultant that you trust in the sense that have you hired someone who's actually good enough that you say, OK, this person, I'd take them seriously? Right? If you haven't done that, you just wasted your money anyway. You know, you got a $10 consultant out of five or something, and you don't trust their response. That's on you. You get what you pay for. Right? Don't do that. Secondly, if you do go with a consultant, that's bloody good and knows what they're talking about and that you trust and you've paid tens of thousands of dollars for them to investigate this and give you a serious report. Why are you not willing to back that? Why are you not receptive to their feedback right now? Trust me, I'm a person who's like in the house. I trust in house because we have the value stream built in, right? That makes sense from a product that makes sense from a long lived investment in a product. But that consultant piece, if you're going to use it right, use it right. Like, actually, you have that investment. Understanding what are you hiring, who are you hiring and how does that add value to you? Right. If you're going to pay tens of thousands of dollars, it's only you as an executive to make sure that there's follow through on that, right? But playing the devil's advocate here, right? Are we then optimize? Let's say I want to flip my logic on, hire the hire consultant, then hire the engineer first [00:24:00] and the data then analyst right.

Harpreet: [00:24:01] Let's take it from the perspective of like that works for a large company who can afford really high quality consultants and then who can afford to build out an engineering team because they're sure of the return on investment in, say, two years time. But take a small team that doesn't have the financial back up to do all of that upfront investment, right? What if you can get a better analyst or a data scientist in to prove the model in in the real world? And you go through this cyclical pattern where your first models are going to be janky and crap, and there's no infrastructure that's properly done for it. And as you go, you've got a cyclically, keep adapting it and improving it and improving it. So. Is that necessarily a wrong way of doing data science? It might not be the gold standard from scratch, but you might be able to build to that right? And is there a use case for going one way or the other based on literally how much upfront capital you can, you can throw into it?

Speaker2: [00:24:54] Love this. Love this. Definitely expanding my perspective on this, and I appreciate you sharing. Let's go to Mexico. You have your hand raised.

Speaker3: [00:25:03] I was looking at the thread and then it was like dressed for the job that you want. And I remember this one. Not at all relevant, but it was hilarious. There was these two guys at this one startup where they were constantly raising each other. This one guy shows up. Talk some more nonsense. The other guy in the other guy is like, you know, that whole dress for the job that you want, because right now you're dressed like an absent father that just gambled away the family savings in Vegas, and it was hilarious. You need to see a picture and you're like, Yeah, that is what that would look like. No, but I guess like, I mean, do you think the kind of one problem, though, in this whole discussion is like, like we're making certain assumptions, right? Like, in some ways, we're almost making an assumption that a business is like totally Data illiterate and there's a hundred percent starting from scratch. And in some ways, they are small enough that they can, like, you know, adjust to the feedback. [00:26:00] And at the same time, they're big enough that they actually have money and resources. So it feels like in this kind of discussion we're talking about like in some ways. Like the perfect consulting opportunity, as opposed to the reality of probably what a lot of companies look like, I guess like. So I don't know. I'd actually be I'd be curious to hear Matt's take on this because I feel like there's a lot of cases where, first off, nowadays, companies are not starting from scratch, right? At least they're AIs or not.

Speaker3: [00:26:34] Sometimes the leadership doesn't always keep up to date, but you know, all companies are not going to say start from scratch like they do have something, especially if they're in the enterprise space. A lot of startups, sometimes they're a little bit too new to actually have the Data to do anything with. And then also, too, like, there's still this kind of like trust factor of the whole like when you hire a consultant. Sometimes people who are incentivized to do differently will not take the consultant's advice, even as good and well-meaning as it is. And that's why sometimes you do need someone to prove like someone needs to literally go. We are losing like a million dollars a year if we don't do this or there's a market opportunity cost of five million if we do do this. So a lot of times that kind of analysis, that kind of like business doesn't come up until someone internally has done the analysis because a lot of times companies will not willingly open the vault of their data to an external party. And if they do, they have to sign a bunch of NDAs. So I don't know, but like I do kind of feel like in some ways, like a lot of the advice of like kind of who we would hire. I do think it is very specific to that company and where that team is at. Like at that point in time. So yeah, that's just like my two cents.

Speaker2: [00:27:53] Let's go to Matt. Let's see. Mexico has high, high praise for your your thoughts here, Matt. No pressure. Hopefully, [00:28:00] I can find the Mubi. Yeah, I guess. I think what Mexico was saying was kind of what was playing around my head. A lot at the very beginning was like, you know, who is the first hire is is really kind of dependent on the company, you know, and we were making lots of assumptions. But like, you know, it's really not a matter of who the first hire should be, but like when you shouldn't make your first hire. Because so like when I was became a data scientist after about two years, I started looking around like I was interviewing, probably like once a month for like two years and kind of Utah area. And Utah had a ton of startups and like data science was this new thing and everyone wanted to do it. But like all of these startups, had no idea what they were doing with Data, right? Like essentially, they're just selling stuff online, you know, just websites, right? And like really a lot of the stuff that people need don't necessarily need a Data team for.

Speaker2: [00:29:11] And so, yeah, it really just comes down to like I got really far in most of those interviews and ended up turning several of them now. And just because the company wasn't ready for really not investment into Data because they didn't have the data, right? So like really, before any team happens, like you need a company that actually has the data and the needs. And yes, I mean, every company has, you know, they have, you know, they need to understand sales data and market data and other things like that. But a lot of that can just be done, you know, with business intelligence. And so I guess maybe that, yeah, a data analyst is all you need at first. But, you know, a data engineer to set [00:30:00] up, you know, Domo or Tableau or whatever is also really useful. But like, can I kind of get a team together to handle all those means? Awesome. And I know Russell, what are your thoughts on this? You have your hand raised as well.

Speaker4: [00:30:16] Yeah. So I've got a couple of responses from the comments since last week and one thing that I forgot this state the last time. So I think I agree well, with Matt, you know, the timing of the hires is very important, you know, so so hire someone. Rather than wait six to 12 months to decide who you should hire first, at least hire someone and start the journey, even if it's the even if it's a suboptimal hire, i.e. you've not hired the right person. So you've gone through the data scientist first, rather than the data analyst or the Data engineer. Whoever is optimal for the situation, at least having some hire and starting somewhere is better than doing nothing. But the key bit that I missed out saying the last time was, I think someone like a Data landscape or a Data ecosystem architect, someone that understands the entire. Ecology of the ecosystem or the landscape is really key, because then they can understand the maturity they can let you know. You know, you're actually quite mature. You could go straight for the Data analyst, honest, the data scientist here. Or maybe you want to start with a data engineer. If they don't, maybe go for too, you know? So there's good insight that can come from that.

Speaker4: [00:31:31] But that role is essential for the for the longevity and the quality of the Data team for its entire lifecycle. Basically, because that's going to be the gel, the catalyst that that brings all of the streams of of Data work together to make to make sure the analysts work well with the Data engineers and the Data scientists, etc.. And then lastly, if we are then talking in realms of large organizations as well, then architects, [00:32:00] then oversight of the entire Data landscape is quite key because it's possible that you'll have different elements or different parts of the business starts and work with data analysis separately from the others little microcosms of data analysis. And they might start with completely different methods, different strategies, different structures, etc. And when they get mature enough to want to consolidate those, they may be incompatible. So having that oversight and understanding of the entire landscape at the start is really an optimal place to be. It may not be, excuse me, may not be the best option for you if you're limited on budget. But if you can do it, I think that's a really key hire to make first.

Speaker2: [00:32:46] So would that be like a kind of like a solutions architect who understands kind of like this larger, larger piece?

Speaker4: [00:32:52] Yeah, I guess so.

Speaker2: [00:32:53] But I'm also labels are needed.

Speaker4: [00:32:58] Yeah. Excuse me. Again, I'm talking about this in the realms of Data only, you know, so you've got architects, you've got software architects, solutions architects and everything. But I think it's important to have one for Data as well. And then those architects can have their own steering group as well so that they meet regularly to make sure that the the overall structure of everything in the organizational ecosystem is also working at the same kind of synchronicity. And there's nothing is going to throw a throw a wrench in the works as it were.

Speaker2: [00:33:35] You bring that up, and that's actually been something that's been very top of mind because in the organization I'm in, we've built up a lot of trust through our work from the engineering team where they're trying to get more hands off and letting us kind of build more so we can move a lot faster. And some of that's been really top of mind is like great. We have a lot of this freedom. We can finally move forward things you want to do. But how do I prevent us coming like a tale of two worlds [00:34:00] where I totally butchered things tale two cities where it is, but two different groups to different systems where they all of a sudden now that connection between the two becomes a bottleneck because we're not, we're not communicating with each other. So I think that's a really interesting point because that is not happening right now, but I want to prevent it because I am taking a lot of lead on kind of this Data infrastructure side of things, at least from the data science perspective, rather from the engineering perspective. And it's challenging because many times we don't speak the same language, so trying to translate between data scientists and engineers. It's been a common theme whenever I show up this week. Just how do I talk to engineers, y'all? Y'all are weird, but we're weird too. So it is weird in different ways. Awesome.

Harpreet: [00:34:48] So I guess this is kind of extending on what Russell was saying is what's the definition of done for for a data scientist or a a data transformation, right? Is it that, hey, we've built models on the data and they run? Is that does that mean we're done here? We've hired someone to build some models. How do you understand that as a business, right? What's the actual requirement for your business to actually see the value that the data holds and its potential value? Right. And this is where someone with that significant architectural understanding that lay of the land of knowing, OK, what's the infrastructure requirements for this? What are the, you know, how can we actually consolidate the data across the whole business? Understanding that topology is quite that's quite key. So having that kind of person there is important. And I mean, yeah, it speaks to what Mexico said as well in the sense that not every company is completely immature and doesn't know anything and are looking for. Please teach me. Right? Like, there are a lot of companies that are quite aware, like quite a ways along in their journey. How do you how do you start by understanding? [00:36:00] Is that sorry? How do you start understanding all the pieces that actually need to fit? You need that kind of significant architect overview of the. Platforms are using here's all the different variations in your Data, right? Really establishing that definition of done is important. Otherwise, I mean, I've seen situations where especially like a C-suite level or senior management level, it's like, Oh yeah, we got the guy to do the model right that was done. Isn't that the data science initiative? And it kind of stops there.

Speaker2: [00:36:34] Yeah, I think that's a good question. I like Russell's kind of point is that the Data work is never done. Oh, so true. The moment that you released Data or a bottle into the wild, it begins to like decay so constantly, constantly hurting herding cats to make sure your Data acts nice. Makiko, you have some additional thoughts

Speaker3: [00:36:57] If you feel like that's something that in the early days of a project is something that kind of needs to be identified. And so is like, what does success look like? Because like for some people, successes of P.O.S., whereas for others it's like successes. The ball is deployed. We have tested it and it's like rolled out in the product. And I do kind of feel like sometimes like this is a little bit hard. So when a project has like a career like product initiative or a product owner, it is a lot easier to say like what is done at that point because the goal is getting into it right. But a lot of times office projects are sort of like so for products are self initiated by like a scientist or just anyone by Nike. Right? A lot of times I do kind of feel like it is motivated a little bit by like shiny pool syndrome, which is, Oh, look, you came out. Let's see. Let's get let's figure out a way that we can use it. And at that [00:38:00] point, there's usually a lot less of a good business explanation for why someone wants to work. But at the same time, a lot of the coolest products that we have nowadays, right? Like, for example, a lot of like Google's cool products, they were sort of developed during 20 percent time.

Speaker3: [00:38:17] So I think there just needs to be kind of like a clear grounding and understanding of kind of specifically like if you have a project, for example, that doesn't have a file donor or even if it does have a product over it, it should still tie it to some kind of like strategic. There should be alignment to the strategic goals, right? But I do think that's like that's I feel like if you had to like ask like five key questions when you first getting started with a data science project like and sort of, you know, like to make sure that's successful, I would be the first one is literally like, what is the what is the output? Is it like an API endpoint? Is it a container that we deploy to like GCP or similar or or whatever their solution is? Like, what is the Data requirements? Because if you don't have the data and there's just no way to get it, then you shouldn't even be like talking about the project. Who's like phoning it like once. It's like once it's been like everyone can agree on who owns it when it's developed. I feel like the the kind of arguments come up when the mall is deployed. So it's like who is owning the maintenance and the operational aspects of that model? So if it breaks like is the data scientists doing like like pushing the new updates or is it like the other one here or whatever, like having that racy chart? I feel like it's so, so important.

Speaker3: [00:39:35] And then also to like having a deadline. And the way I look at it is like, if you don't have a deadline, then it's probably not important to the business because that's that's that's honestly where you get like. To me, it's the most annoying thing to like. See, like the business partners come back and be like, Oh, we need this in a month and I'm like, I'm sorry. Like, I'm not here to pay for the fact that you weren't on your [00:40:00] on your Shahnaz. You know, you weren't doing your business and letting us know where the project was going. But at the same time, that's also a way that you can then sort of extort some sorry, you can get some good kudos for the future. So, you know, maybe around promotion times, everyone should have that promotion notebook that package, right? Just go do a line on this day. This scrutiny over and no nice save the project. Everyone has a bad book, so that's sometimes some things are worth like owning. And then you put in the brag book and other times it's not worth putting it down in brag book.

Speaker2: [00:40:44] Yeah. Well, let me move on to thanking you for sharing your thoughts. We'll move on to the next question. We have Sarala, who I believe is a new guest for us. You had a question, so I'm super excited to kind of hear what you have to ask. Give a quick moment,

Speaker5: [00:41:05] That moment you take a bathroom break.

Speaker2: [00:41:07] Yeah, we can, we can go back. Eric, I know you had a question.

Speaker5: [00:41:12] I did so. So I've been working on this project and I I showed it on LinkedIn. It was my like Murphy Index or Irish name scoring thing, right? And so Harp, you had an interesting idea. You asked if I was like collecting any of the data and I was trying to think like what I could do that would make it that would make it worth collecting anything, you know? And one idea that I had was it could be interesting to collect, even if I didn't collect the names of the that were submitted, just collecting the scores and to be able to say, you know, for future people, you are in the top 10 percent or here's what the distribution of scores looks like or something like that, right? So that's kind of the first idea. So my question was from [00:42:00] what from a, you know, GCP or a like in terms of that, it's just a streamlined app right now, and all it does is run the little calculation based on like the pickled Data that I have in the in the code. And so what would I need to do to take that and save it somewhere and call it again to be able to, like, visualize it? What what are the word pieces of the platform that I would need in order to set that up?

Speaker2: [00:42:33] So that's definitely outside my wheelhouse. I think it's more so going to the mop side of things you're asking about where to store your models and whatnot. So I'm going to look to Mexico, who's on this side, at least for me.

Speaker3: [00:42:48] Data Engineering That's in the house now. Mr. Sharp, actually for what would be good data storage solutions.

Speaker2: [00:42:56] Um, I'm not too familiar with streamlined, but.

Speaker5: [00:43:01] Okay, so it just when somebody types their name in, it just saves it as a variable and then does stuff with it. That's all like super simple,

Speaker2: [00:43:10] Like, is this running on the cloud or a server or what exactly? It was like a really streamlined share. Like, it seems like this is just a very simple application that I would probably just say it to a text file and then log into the server and download it. And then if it's not something that is being saved like it's in some container that's being destroyed, if it's not being used, then then I'd probably like push it to like just a Google Sheets and then go from there and figure out maybe how to connect it to something more permanent from their upload. If it saves it to a CSV file, upload it to like S3 or something some [00:44:00] storage object storage.

Speaker3: [00:44:01] So yeah, GCS would be like the equivalent of S3, so it would still be like appending to like a text file or something.

Speaker5: [00:44:09] Yeah. Okay, so first grade question to your first to your first grade solution when you say, oh, just save it to a text file, like, can you say a little more about that? What that looks like is it like, am I going to need to? So as on, so streamline share just plugs directly into my GitHub repo? And so where does that text file live and how do I connect to it? Or yeah, where does the text file? I can figure out how to connect to it and do stuff with it, but I don't know where that stored.

Speaker2: [00:44:39] So quick note. Oh God. Like, it could just live in your GitHub, right? So like, you can have text files as inside of your repo and it can be and stored as LFS, which will have some limits. But. And you can just kind of work together until you come up with a better solution.

Speaker5: [00:45:06] So somebody goes, let's say that you go to the app, you type your name and sharp, so it's going to grab that and then push it to like it's going to when you load it, it's going to like pull the text file or pull the repo or something, update the text file and then push it back.

Speaker2: [00:45:26] But yeah, so if you're if you store a file and GitHub with lrfs large file storage, it is your GitHub repo is literally just saving essentially you URL and pointing to the storage, and that file is being stored in some cloud storage similar to SARS in the back end. And so then you can just constantly update the actual file and then inside the GitHub, it'll [00:46:00] just keep on pointing to that source since it's just a pointer. Does that make sense?

Speaker5: [00:46:05] Yeah, okay. I don't know if this was a well, I didn't know the thing. And then until five seconds ago and I didn't realize it's like a separate thing. So this is going to. Ok, cool. That's really helpful. I can.

Speaker2: [00:46:16] And honestly, this might be like a terrible solution overall, but it's it's just like, I have no idea what you're working in. So this is just throwing out ideas. So also real quick shout out, shout out to LinkedIn. Paul Fentress is saying that Tremlett has a Tremlett Cloud option to save your data, so I tagged you in there so you might want to check that out. There may be something already pre-built within streamline it to make your life easier.

Speaker5: [00:46:47] Oh, that'd be awesome. That must be why they got purchased by Snowflake.

Speaker3: [00:46:50] I was going to say that's where your monitor didn't have it. It's the Data story. That's what they're going to be. Yeah, that's that's where the pricing is going to come in. It's like the number of notebooks, instances and data storage ready.

Speaker5: [00:47:05] Cool. That's helpful. Thank you.

Speaker2: [00:47:07] Awesome. So let's go back to Sarala. Let's see if you're available to ask your question. Hello.

Speaker3: [00:47:18] It's my voice audible.

Speaker2: [00:47:20] It is.

Speaker3: [00:47:21] Oh yeah, oh great. First of all, thank you, Eric, for sending me an invite to the session. And it was great meeting all of you here, and I'm really happy so that, you know, at least I can post one of my questions here. So my question is I am a sequel Data and the list of I do anything and everything with sequel. That is what I do day. And I'm an occasional Python developer as well. So right now, I'm at a phase where I feel that I'm kind of stuck in with the sequel and the Python. I would like to do more. I mean, like, I would like to go beyond the sequel and Python Analytics. [00:48:00] So I thought maybe venturing into data science and machine learning would or would be the best next option for me. But then I'm not really sure where to start when I look at the plethora of options that that we have right now. I mean, it feels quite overwhelming. And I thought, maybe you know, I can ask this question and get your inputs on this to see, you know, what could be the starting or the very first point that I can start off with for this journey, for my journey into like I'm actually trying to build a good profile so that I can start applying and make that career transition. So if you can answer my question, I think it would be very grateful.

Speaker2: [00:48:47] So I'm open the floor to anyone has thoughts on making this transition from being strong and sequel knowing how to go Python apps, but moving towards a more data science side. And imagine people are going to ask more questions first to get more clarification, just to learn, learn more about your experience.

Speaker3: [00:49:07] Gina Yeah, just real quick as somebody who's also transitioning to Data science, although I did a bootcamp and have been working on other projects, the fact that you've got Python in SQL was huge because pretty much every single thing I've heard, including from mixing in his book and in his interviews like Python and Sequel, know these for pretty much any data science interview, right? I mean, just be prepared and you probably get asked about it. So that's that's thing one. So take heart in that.

Speaker2: [00:49:46] And I would echo those when you said SQL, I was thinking to myself, I'm like, I think like 80 percent of my job right now is skill because I use it just to pull the data you can do to the stuff. So you may switch jobs where you still might, might doing very [00:50:00] similar job in some kind of way. Mexico.

Speaker3: [00:50:04] Yeah, I'm good. I mean, we have plenty people on this call, I think are better equipped to provide a good answer. But I think I would always sort of ask yourself what you hope to get out of your next career transition out of your next job. And the reason why I kind of say that is because. You know, it's easy for a lot of us to kind of throw like a list of resources and skills at you, and we all have our own different opinions and experiences from like our kind of career trajectories. But I think it's important to kind of know your why. What kind of projects and what kind of work do you want to be doing? What kind of team are you looking for? The reason why I say this is because, you know, a couple of days bootcamps, I sort of mentored that there was a portion of people for whom they kind of saw moving into data. Science is sort of like the next step up the ladder. And it is a different type. Working as a data scientist is a different type of role. There are skills that Venn diagram. There are skills that overlap. Absolutely. When you think of science or when you move to work as a Data in here or in an office or what have you, there are other skills that you will have to pick up. It just it is what it is.

Speaker3: [00:51:23] And part of being successful is understanding that gap between what you have now versus kind of where you need to get to. But it's also like kind of understanding your why, what is the type of work you want to be doing? What type of products do you want to work on? Because I think the trend nowadays is that in a lot of different types of roles, it's kind of expected that people have some experience with machine learning models, even simple ones and have some kind of understanding of it. So for some people, like at the students who can be mentored for them, what they just want, what they wanted was they wanted. They want to grow in their career. They want more money. [00:52:00] They want want to have more strategic role. And so they learn machine learning, but they learn machine learning to become more senior Data employees who are kind of able to sit at the table to inform on projects. Some people, they, you know, they still went into a data scientist role. It was a little bit closer to the way a data scientist role, like a proxy analytics role thought of that some companies like Facebook and Apple and all that. Some people went to more of a research scientist role, but a lot of times the kind of the role that you want to go for and like the skills that you need to then sort of build up to get to that role, it's dependent on the type of work you want to be doing.

Speaker3: [00:52:42] So I would always kind of like, ask yourself your why and kind of what you want to get, what would be that? And then and then you can kind of like map the skills a little bit more more easily because a lot of our data scientists, for example, yes, they are building models. But and this might not be the case at other companies, they're building models that will get into production. So they do need to know some engineering skills, but also they're working very closely with product. And so a lot of times they are also developing that kind of like, I don't want to see deep domain knowledge like you don't have to know one hundred percent commerce to work on E! Commerce App. But if you're working on a recommendation out for an item to go into like a cart, it's probably good to just understand. Like, OK, like this is like, this is the funnel that products go through. These are the metrics that they care about all over stuff. So that's just kind of the thing I would sort of put out there. But other people can kind of talk more about the actual process, like Eric Sense on how do you serve to become like a really successful, successful analyst and partner? And Mark two and that show up here.

Speaker2: [00:53:57] So we have start next. But I'm curious because you called out Eric [00:54:00] Sims if Eric, you have any quick thoughts. If not, that's also completely OK as well.

Speaker5: [00:54:08] A quick couple of quick things related to what Michael was saying, like just being a cool person to like, talk to and whatever goes a long way in opening up opportunities that you're interested in. And then just kind of that list of resources stuff, it's like, how do you get involved with it or get started with it? One hackathons, because then you can work on somebody else's project that's definitely further down the road than your project. That's an empty Jupyter notebook right now. That's helpful. And then the other is just like I try to find things that are going to happen that helps motivate me. Like St. Patrick's Day is like, Okay, I can do a project about St. Patrick's Day, Halloween, I can do a project about Halloween, and it just makes me think about, like, help me guide my ideas because otherwise I'm just all over the place and I don't know what to do and and I'm just never going to start, and I just have too many, too many ideas. And so like narrowing it down and just having something and then digging into it, you know, you hit enough roadblocks to like, learn, learn lots of stuff along the way. So anyway, that's my quick two cents.

Speaker2: [00:55:14] I'll also add kind of the putting a holiday next to also time boxes, you know, like I have done before high. Very motivating, because like do a Halloween project on November 5th, you kind of missed the mark right there.

Speaker3: [00:55:28] No one's asking you to kiss me because I kiss me. I'm Irish. No one, no one's doing that. Like the week after like St. Patty's Day.

Speaker5: [00:55:36] Exactly.

Speaker2: [00:55:39] Then has other thoughts. But stop.

Harpreet: [00:55:43] Okay. So I mean, I was essentially going to say something similar to what Eric was saying. Where? Yeah, sure. Everyone's going to tell you to go build a build a portfolio. Right? Easy to say. Easy to find resources. You can look at every medium article under the Sun that says, Oh, top 10 [00:56:00] portfolio ideas for budding Data scientist, right? Go for it. Knock yourself out. You're not going to stand out, right? Pick something that you're deeply passionate about and do a project based on that right. Next thing is a great example of this. I think from his story, it was the R&B something to do with R&B music. I think it was something along that line, right? Do something you're deeply passionate about. The rap was the rap music, right? Okay. Yeah. So pick something that you're deeply passionate about that's going to set you apart, right? Maybe you're the only person who just did it. My data analytics model, UN kebab shops. All right. His kebab shops are great. So like, set yourself apart, like, that's what people are going to look for. That's going to say, Oh, well, this is this person.

Harpreet: [00:56:47] As opposed to the one hundred different resumes that came along that did the same project of, you know, Medium.com June 2020 article, right? Set itself apart. But I mean, they kick off the big question here, and I don't think I mean, I don't know if you're willing to or able to answer that question necessarily right now, but why? Like, for me, it's very clear why am I in data science and machine learning? My mission is I want to bring robots to the real world. One of the big four challenges of that is making them sense the real world. So to me, my mission is how do I teach robots to see? That's the bottom line. Everything I do takes me one step closer to that, right? And sometimes I've got to take a sidestep and figure out how cloud works to figure out how to run it on robots. But that's my big mission. Do you have an answer to that? I'm really curious to hear, sir, if you actually already have a cohesive answer to why.

Speaker3: [00:57:46] Um, for me, as a child, I was, you know, I was always amused by the fact like, you know, when I would, you know, scroll through YouTube and see, you know, it would, it would give me some of the recommended videos, right? At [00:58:00] that time, I really did not understand the science behind it. I did not know that it is, you know, there is a device that is tracking all my, you know, the kind of usage and whatever I'm watching and whatever I'm seeing. And and based on that it is providing me recommendations is something that I was not aware at that time. And then gradually, when I actually started working on my first job for me, SQL was just like any other subject and during that time, but after I started, after I started writing queries and after I started noticing the kind of change it is bringing to the business table because I would work with school to tell the team, OK, so this is how your product is working pre market sales and this is how it is working post sales. And you know, if I am able to bring out such kind of insights through SQL, then I wanted to explore, go beyond the SQL and see what exactly I can do with the data that I have.

Speaker3: [00:58:55] If I have to give you an answer right now, it is. It does just that I wanted to do and understand greater things with data. And for me, with the kind of knowledge that I have and the kind of understanding and the perception that I see around, I feel that probably, oh, you know, email or Data or Data engineering would be my would be my know the very next bet that I can take it on. So for me, it is always leveraging the data that we have at hand to bring out the best to the to anything that we have around us. So that is my answer, and I am actually trying to try trying to find ways to get there. But there is this constant feeling that, you know, whenever I go and check on the LinkedIn, I see people doing many things, but in front of them, I feel very little or less. I don't know if I can, if I'm putting it right correctly across. But yeah, I feel that I I missed in the trajectory that I have to get it and get on to achieve my goals. And that's how I once I reached out. Eric and Eric had brought me here. So yeah, so I could

Speaker5: [00:59:59] Say something [01:00:00] about LinkedIn. Oh, go ahead. Ok, so I love LinkedIn Total LinkedIn junkie, but one thing I will say is like, I can't I can't let myself read too much into so many things that people post, because when I post something that is like something that I learned or something I'm stoked about, like I learned some stuff, some really interesting stuff about like, compliment, naive Bayes. But if I write a post about that and nobody going to see it and nobody going to like it and nobody going to care, but it's like something that actually matters professionally and can actually be useful to me. But if I get a certificate or 100 percent on who knows what quiz, you know, that's just going to like, light it up and make it look like, Oh, so awesome. So like, you know, grains of salt, take them as needed, as many as needed whenever they're needed.

Harpreet: [01:00:47] So and I'll just chime in there. I'm just going to chime in that I'm a cereal. I don't post on LinkedIn kind of person, right? I will like I will comment on other people's posts, but I am always dead scared to post what I've learned this week, because what I've learned this week makes sense to me as shit. That's cool. I needed that for my personal development, right? And that's big for me, right? And it might be something super basic to everybody else, and I'm sitting there going, If I share this, people are going to be like, Oh, it's not all these learning kind of thing because you only ever see the like. We're only servicing these people. They're like, Oh, I finished this entire Cowgill competition, got a medal in it. I did this massive thing, right? You only ever see the top of the top. This is like you've got effectively the Instagram model equivalent happening on LinkedIn, right? You've got your LinkedIn Data science. The models, right? And it's it's intimidating. It's not easy. And you're like, cut out the noise mod like for your own personal like self self-worth and you just sanity cut out the noise focus on. And this is why that why is really important.

Harpreet: [01:01:57] I don't care that someone else has [01:02:00] become the genius and graph neural networks, because that doesn't help me teach robots to see that's my grounding feature, right? This is why it's not just about what are the techniques I'd like to work with. It's not about the tool, it's about the end goal, right? If you have that vision in mind, I keep going back to that. So I don't care if I don't put us on LinkedIn. I don't care. What I've learned is something so dumb that any infrastructure or cloud architect would sit there going, Oh man, this guy's stop at its job, right? Because I know ten other things that they don't know. They're regard to serving on robots, and that's my mission. Why are we measuring ourselves to what other people are posting so. So look, I hear what you're saying. Don't worry about what other people are posting, right? In terms of all, they're moving ahead. They're moving so fast. Am I missing out FOMO? Figure out why you want to do it and then cut the shortcuts. Don't don't waste your time learning stuff that other people say is important, right?

Speaker3: [01:02:55] I do think it's important to like, have like literal blackout periods on social media. So like when I was doing when I was seeing like the Data Science Bootcamp like that is sort of kind of the nice thing about structured environments is that part of what they do for you is they kind of they sort of give you that that sort of goalpost like they give you that thing to like the outcome to measure yourself against. Because like sometimes self learning is a little bit hard. You have to kind of like, set your like, you have to kind of know and set your own milestones. And I feel like it's hard to do that if you haven't already gone through that path already. But for me personally, like I'm very, very obsessive about planning everything like it's actually kind of destructive, sometimes like how much of my life I live in spreadsheets, but it can kind of also help to when there's a lot of that. So something that does help for I think LinkedIn is to understand, kind of like the landscape of like who is posting and what and why are people posting there? So number one, people are posting [01:04:00] because personal brand, it does help, you know, it helps set the narrative for recruiters and all that. There is some apps talk going, I'm missing out, you know? So that's one reason why people post it's personal brand. You know, there are a lot of people who have, like a lot of accomplishments that don't post, and there are people like me who have like, you know, who don't really accomplish much and we just talk a lot.

Speaker3: [01:04:27] You know, the second like group that you see is a lot of like vendors and, you know, a lot of creators and all that who are kind of selling products and, you know, all that like, there's a couple of us on here, right? It is what it is. We all got to make a living. You know, the third part is also it's potentially people like around you also who are like searching for jobs and it is hyper competitive. So I do think sometimes it what really helps is. Ok, so two pieces of advice I hear given to new people that I really think is B.S.. Number one, follow a bunch of like people and newsletters and. Losers and all that. I actually think that's a terrible advice when you're new, like, don't do that. There might be like one or two newsletters I would almost recommend to a new person, mainly because like, there's this one newsletter where all they do is they list three blog posts, three projects and three papers. That's it. And they send weekly like, I think that is very consumable for a new person. But anything more is really bad. So I actually would not even do that like so as a new person, if you're like getting into a job search and you're actively, for example, try to build on your skills in your portfolio.

Speaker3: [01:05:34] I would just cut out a lot of that stuff. I do like social media blackouts. The second part is like, you should work on projects, but you see a lot of people like their GitHub repos like they have, like all these like like millions of repos and code gists. So I found out that like what some people apparently will do is they will just like fork something. They'll make a little change and they'll call that a project, or [01:06:00] they'll just do collections of little code snippets, right? Like that's, you know, a law that's like vanity metrics. Hiring managers, when they do look for portfolios, they are, they're actually OK with fewer projects that are just really well done like they really are. I have there is a few students like who the the mentoring program like. I maybe I can go track down their projects, but they literally have like two or three repos. They're adding more, but it's just really well done projects, you know? So there's also that where it's like you could kind of drive yourself nuts by like trying to do so much and compete on the numbers game. So where you can compete is quality. It's quality, and it's putting on your own unique spin. The the thing that so the third thing that like the advice is given to new people that I don't really love is, they say, reproduce papers.

Speaker3: [01:06:55] If you're going to do a science machine, I honestly don't do that. Like unless you're going into like a research area, you know, I wouldn't really do that. It's nice. But the reality is that like, for one thing, we do want more papers to have code attached. That's the real thing. Write papers with code. So if you are interested like you can take a look at that website. But what I would actually think, for example, that is a little bit more valuable for a data scientist that I don't see in a lot of science, frankly, is improving like sort of basic software, ensuring skills. That's actually, like, really, really powerful. And so a lot of times what's probably better is to like, find a good blog post or write up of a project someone did and then really kind of go through it and reconstruct and diagnose or dissect every the decisions they made to essentially get like a a model up and then to get it sort of like. How did they how did they share the model with people? It could either be through like flask, it could be through streamline it, it [01:08:00] could be through whatever. And like, how well does the model run like if you download and pip, install it on your laptop and it doesn't work, then I would argue that's not a successful project, right? Like a successful project.

Speaker3: [01:08:11] Hypothetically, if you do everything well, you know you set up your virtual environment, you pick and sell or whatever. It should work, you know? So but that's more what I would kind of try to focus on, like really cut all that noise. Don't do all the following. All the influencer stuff like really restrict kind of your work and learning to something that's like, well, encapsulate and high quality, you know? And then eventually, when you get there, like, share your learnings on LinkedIn, it's kind of only scary if you sort of feel like LinkedIn is being done to you. But if you are a participant and LinkedIn, if you kind of build up that community around yourself, it can be a much more welcoming place. So that's I know a lot of words. See what I say. I talk a lot and no skills at that know me. They don't meet in my bike, but or my bark is no bite, my bark, whatever, you know. But that's you know what I would say, like for new people. I do think there's a lot of advice. There's a lot of noise out there that honestly, I think it's just really not great if you're new and you're trying to pivot. But quality quality is never a bad thing, especially in a very crowded field. Thank you, thank you for the perspective, yeah.

Speaker2: [01:09:25] Awesome. Did we answer your question then kind of chime for these next steps for yourself?

Speaker3: [01:09:32] Yeah, yeah. I think I got a very good perspective on what I'm looking at right now. Maybe I'll pick a few points from this and start building up the road map to where I want to be. Thank you so much for that.

Speaker2: [01:09:45] Amazing. Awesome. So. Well, we might have time for two more questions, but our next question is Gina regarding some resume work. So going back to the kind of like a job [01:10:00] search and LinkedIn kind of thing?

Speaker3: [01:10:02] Yeah, I think so. Speaking of advice and a lot of noise out there, one thing I mean, obviously there's tons of resume advice and there's resume advice that might go for more of a project manager program manager outside of Data science, just for one example. And then there's Data science resumes or software engineering, etc.. One question I had is so a while back, some years ago it was said that you shouldn't do any formatting on your resume, for example, any boxes, any graphic elements, things like that, columns, et cetera, because it won't get through the applicant tracking systems. So I have one question around that to start with, and I've heard various things. It seems like you should have a skill section, but I've even heard some say, don't even put that in there, but try to capture if you worked with SQL or Panda's Python, all the rest capture that in your technical projects section. So I'd love to hear people's thoughts basically about getting through that. That's obviously the best thing to do is to try to connect with people who are hiring, et cetera. But and also this is, I'm thinking, UC focused. You know, I don't know how things are in other countries as far as what is expected from resumes, but here it feels like, you know, one page Max unless you're extremely experienced in this particular field. And also, you know, oh, make it visually interesting. Well, that's nice. But if it doesn't get through an applicant tracking system, then I don't know that anybody sees it unless you actually [01:12:00] are talking directly to a recruiter who walks it over to a hiring manager, et cetera. So love to hear your thoughts on that.

Speaker2: [01:12:07] So let's go with Matt first. Oh, yeah, so I've probably helped thousands of students with resumes like I've been through the song and dance a lot, like whenever you give advice, like everyone disagrees or agrees or, you know, like should it be, one page should be to page. But should this be, what should that be like? A lot of that really comes down to style and how you want to preference yourself and like you really want to pick something and lean into it, you know, like you like. No matter what you do, you're going to impress some people and you're not going to impress other people. And so I generally like I pick what I think will impress the maximum amount of people. And so like so I generally always keep my resume to one page, even though I could definitely do two, three, four pages, you know? But I keep it to one page because in my mind, like people just scan them really quickly and then like, really, it comes down to the content, right? And so like I often tell people throughout the skill section like you were saying, but that's because like it all comes down to like if you have amazing work experience, you want to highlight that and you don't want it to be buried behind the skill section, that's just taking up space. You know, obviously, if you were able to do this amazing project, you obviously had to know the skills to do it, you know, probably, you know, if you deploy the GPT three model that did this great thing.

Speaker2: [01:13:57] Yeah, you got to know Python, [01:14:00] you got no API, you got to know a bunch of these skills. So like, it's kind of already inherent, but like really the only reason the skill section is in there is so that way you can like bombard these, you know, these automated systems that just try to like, filter you out automatically. So like that way, they can automatically know the the words you're looking for. But like you can figure out like those automated systems aren't smart. Like, if you want to know how they work, they just take the job description. They take your resume and they word compare. And if if you have more words that match and meet a certain threshold, like 10 percent or whatever it is, then you're golden. So like if you want to get past an automated system, all you have to do is look at the job description and then just throw those words into your resume and then you're good. And then once someone sees it, then they're going to really care about like you. In my opinion, you should always write your resume. So that way, when a person sees it, they're impressed, not when some computer systems use it. Like you should never care about the computer system. In my opinion, if you get filtered out at that level, then you weren't a good fit for that company anyways. So. So those are some of my strong opinions, but that I say lean into and you don't have to agree or anything, but those are just some insights.

Speaker3: [01:15:27] So follow up is simply so yeah. And I've heard this many places as well. But yeah, it's been really impressed upon me and I've actually been working with a coach who's been doing this. She's a wizard at stuff like this, and she's great. Not all kinds of other ways, but getting those matches. So if a job description calls for, let's say, Tensorflow, which I've used in a project and you know, I could learn and pandas and python and terrorists sort of Tensorflow ish, et cetera, [01:16:00] et cetera, then I want to make sure that those show up in my project description somewhere. So if I don't have that skills section, then I want to at least make sure that I'm getting not just those words, but other words right from the job description of similar kinds of words that will help it get through the system. And or if it goes in front of a recruiter, it makes it real easy for them in a quick scan to see, you know, to kind of go, Oh, I can see that this person like even if it's words that aren't technical skills, but it's kind of wording things in a similar way, you know, that just makes that comparison easy. Would you agree with that? Because on the one hand, you're saying, you know, write it for a person, and there's the thought that someone mentioned, maybe just say, if you've done these projects, it's kind of implied that you know, these skills or you have these abilities and yet you're trying to get through that. It's then like you said, the computer isn't going to know that. So just clarifying on that.

Speaker2: [01:17:09] Yeah, I mean. Than what you're trying to get through, right? If you're trying to avoid the system altogether. Like if you have a referral, you're probably going to ignore that quality system altogether, right? Yeah. And so then, yeah, just just optimize it for how you write it. But if you if you're applying to some, some big company and you don't necessarily have an in and like, you know, other things like that, then yeah, like you do need to get past its system and go through and kind of maybe, you know, take the job description and sprinkle in some of the key words into there. But you know, ultimately, like a well-written resume, like you'll notice, Oh, I already have all these write words in there.

Speaker3: [01:17:55] Ok. Oh, and one other thing formatting. What do you know about like, [01:18:00] literally back in the day, it was like if you have a line in there, if you have like a gray box or a text box of some kind that can get filtered out or even just which format I heard people say use PDF only. And then I've heard people say, you know, use a Microsoft doc. And if it doesn't say in the application system, which sometimes it does not, then I mean, those are just these are like little silly technical things. But it kind of would suck if you are applying for something and putting some effort into it and then you just get screened out and you don't even necessarily hear back or know why.

Speaker2: [01:18:38] Yeah. So the format like you'll know when when you go into play, right? Like if all you have to do is upload your your resume as a PDF and that's it, like, then you know that they're using some system that's extracting the information out of your out of your PDF. And so then you're format. Maybe you might want to do a more simple format because because then you don't have to worry about whatever system is screwing up, extracting that information out. However, like still today, like they ask for your resume and then they ask you to fill out the form. And this happens a lot and people always complain about it. But like, if that's the system you see, then you know that you're good. You can upload a resume that's very visually appealing, that has whatever format you want because you're going to be copying and pasting all this information over into a system that will then be able to extract it right. And so. You just kind of got to know what system you're dealing with and a lot of those systems that are extracting the information out of resumes or getting a lot better. And and so it's always pros and cons, right? The like ultimately like you always [01:20:00] want to try to get around and like you can always like if you can reach out to a recruiter and ask them about the system like they're always happy to help you, right? Like, like if you're an amazing if you're an amazing candidate, like no one wants to lose you because some system filtered you out, right?

Speaker3: [01:20:19] That's right. Yeah. In fact, I was advising somebody on that the other day, but it was specific to where I had worked before UC Davis, which Mark knows well. All right. Yeah. You know, I hear these stories about the resume goes through shared services. And in fact, somebody had reached out to the place where I used to work and they said, definitely apply. And then they got filtered out right away and nobody was happy about that. And in fact, it was, I think, the HR people who were like, Oh, we didn't see this person's resume. We got to get in touch with them or get in touch with shared services to get them to send it. So that is and this is just like something I kind of picked up on, and I think it might be helpful for others who are listening or who tune in later to kind of plus one that and say you're helping to solve someone else's problem. And if you reach out to them, that helps a lot. And also some notes that Mexico was saying in the chat about she has a resume and an in-person resume. But she says she often uses the the resume because it's real, clean and, you know, and sharpened to the point. And as somebody who has been known for worthiness in my past, I've come to also appreciate like signal to noise ratio. If you maximize the the quality of the shorter resume, that one pager or anything you write, then it's going to have more impact than if people kind of have to wade through a bunch of stuff. [01:22:00] So, yeah, thank you for that. Thanks so much, Matt.

Speaker2: [01:22:04] And then we're coming up on time, so we've got time for one more comment. So what you have to share?

Harpreet: [01:22:11] Yeah, I mean, I kind of moved the other way in terms of skills versus experience, right? But essentially, it depends on on you if your skills are the things that are going to get you the job, but the emphasis on the skills, if it's the experience, it's going to get you the job, put the emphasis on the experience right. If you don't have the experience, put the emphasis on the skills and projects and flip it around when once you do have the experience that doesn't, then I'm going to ask you whether, you know, Python or Tensorflow. If you worked for 10 years as a senior data scientist, I don't think they can really ask you either way. Right. So you get so focused on your strength. That's the first thing to know is what is your selling point? Is that the skills or is it the experience? The second thing is who you are applying to is that a large company that's actually going to use one of those ecosystems or its systems? Or is it like a 10 man group? Are they getting thousands of applicants or are they getting 20 to 50 applicants, right? If they're getting 20 to 50 applicants, they don't want to waste their time setting up big apps, and they're probably going to go through each one anyway. Right? If they've got thousands of applicants, totally agree. You've got to be careful what you send through just whatever tool you do use, make sure that it is readable from it. Like, I know that using funky fonts is usually where PDFs sometimes encode things in a weird way. I know Canva does a few things with their with their fonts to there's some of their proprietary fonts become hard to read. I found one at a system reads my name backwards, but let's I've been in that weird situation that I've had a teacher recruiter reach out to me saying, Hey, your resume broke our ATL system, but we really liked your resume. Let's have a chat, right? Which is the most bizarre experience I can have a comment on. But that's a

Speaker3: [01:23:59] Strategy. [01:24:00] You break the Attias and then someone has to look at it like, you know, a clogs. It's like a paper jam, so to speak. Oh, it's

Harpreet: [01:24:08] Not my highly recommended strategy, but just not one thing. Like, Yes, I'm a believer in visual resumes, right? It helps a lot, but essentially I'm throwing back all the time to some of the stuff, Vince said on one of his videos. Is your resume is the thing that you spend a lot of time creating, but people spend very little time reading, So how do you get the minimum effort to get the resume that matches the one hundred different jobs that you're trying to do? Value your time, man. You don't want to spend four hours writing a resume, so sometimes keeping it simple is good. Or once you have a visual template and an ATR template, just update it and move on. Don't reinvent the visual every single time I spent 22 versions. 2019, which I think 20 of them were a waste of my time, so don't do what I did.

Speaker2: [01:24:57] Awesome. Well, we're at time now again. Thank you so much for everyone showing up and sharing this awesome conversation and big shout out to Harp for trusting me with hosting your show. I'm so excited. I'll chat with you all nerd out about Data. You know he has his phrase. The one life thing. I'm bad memorizing things. So like, I'm just going to butcher it. But I will leave with this is that I live by in every seat is a teacher and every seat is a student and we're all learning from each other and everyone has the opportunity to teach someone else. So I feel that every time I show up here and I really love hanging out with you all these Fridays really help out my career and also hope that your careers have moved forward as well from being here. All right. Have a good weekend, everyone. See you.