HH68-11-02-2022_mixdown.mp3

Harpreet: [00:00:08] What's up, everybody, welcome, welcome to the artist Data Science, happy hour. I was about to say at this hour, just tell you how crazy my week has been. But thank you all for joining me today for the artist Data science. Happy hour! It is Friday, February 11th, 2020, to hopefully get a chance to tune into the episode that I released today. If you haven't, please do because it's a good episode. Speaking to a Lisa Simpson have a hard time pronouncing her last name is Rose Schroeder from well, she's author of a book called Real World A.I. It's a great book. We talked a lot

Harpreet: [00:00:42] About product

Harpreet: [00:00:43] Management for data science and machine

Harpreet: [00:00:46] Learning, and how to structure data

Harpreet: [00:00:47] Science, machine learning teams and some controversial takes on who the first person she would hire for data science team is. So definitely check that out.

Harpreet: [00:00:56] Great episode.

Harpreet: [00:00:57] Also, hopefully got a chance to check out the comment. My office hours had a chat about all about machine learning reproducibility with with an internal person, the head of research at Commente, Dr. Doug, and also with Tiffany Fabiano, who is from AstraZeneca. So it was a great, great conversation. This upcoming week. We're going to be doing just a solo session me for the first half of the office hours, just kind of given a lecture style thing and then also got

Harpreet: [00:01:27] My colleagues

Harpreet: [00:01:28] Drew and Michael coming in. And we're going to be talking about the work we've been doing behind the scenes for the working sessions. So please do go ahead and check that out. Man, it's good to see a packed office hours or happy hours session today. It's a lot of a lot of faces I ain't seen in a very long time. Jennifer, what's up then? What's up? Rashad what's up? Marc Freeman is in the house with a special guest, which we will see as soon as Mark returns or terms on the camera. Shout out to Alexandra A.

Harpreet: [00:01:58] Russell can have all the guys [00:02:00]

Harpreet: [00:02:00] Here. I'm super excited for a four, for all you guys to be here the next two weeks after this, this session will be hosted by Antonio. He's going to take over while I'm visiting friends and family in Sacramento and San Francisco. So, Antonio, thank you for taking over. So let's kick off a question. I want to. I want to know, man, I want to get you guys to stars on this ageism in Data science. Is it a thing? Is it real? Is this an issue that we're facing ageism in Data science? I'm curious, man, because you know, I'm pushing 40 and really, I've only been in the field as a full on Data scientist for three years, even though, you know, I was doing quantitative stuff for for for many years before that. But I'd love to give people, is that people's takes on this ageism in Data science? Let's go to let's go to Vin and then let's go to

Harpreet: [00:02:55] Tom after that.

Harpreet: [00:02:57] And then whoever else would like to jump in on this, please do anything. If you're watching on LinkedIn, if you're watching on YouTube, if you're here in the in the in the session, please do let me know if you've got a question. Drop it right there in the chat. So yeah, let's kick it off with first with that, Vin. What are Tom that maybe we could hear from Alexandra? I mean, we can hear from Rashad. I just love to get your get your thoughts on this ageism and Data science. Shout out to Morgan Freeman in the building with the one and only Jeremy Jeremy from NACI. Yeah. Let's go to it. Then let's say let's hear from you.

Speaker3: [00:03:34] Yeah, I think it's definitely a thing, but ageism in Data science is different than ageism in software development and all of the other technical, it's like we get another 10 year grace period instead of, you know, everything kind of sliding at 40. Everything starts sliding at 50. But outside of that, it's the same. It's the same ageism. It's the same. It doesn't make any sense because you need people [00:04:00] with 15, 20, 25, 30 years of experience to be on your team, especially at those top levels when you want them there, when you're doing a new type of project,

Harpreet: [00:04:12] Or when you're

Speaker3: [00:04:13] Trying to implement more advanced methodologies.

Harpreet: [00:04:16] There are so

Speaker3: [00:04:17] Many different reasons why you want someone who's been through multiple cycles like I've been through software, SAS Cloud,

Harpreet: [00:04:26] Big Data, Data Science.

Speaker3: [00:04:29] You just want someone that's been through each one of those cycles. And as you're starting to ramp up and build out solutions, you know, people like that are able to interject some sort of logic and reason. But the ageism is definitely there, and it's really a

Harpreet: [00:04:47] It's a function of how fast

Speaker3: [00:04:49] We have to promote people to management and leadership in technology. We have a whole lot of leaders who are in their 30s, and if you go to other organizations,

Harpreet: [00:04:58] You don't see

Speaker3: [00:04:58] As many 30 year old veeps and you don't see as many, you know, really people that are at that age controlling the hiring and controlling the perception of who should be brought on board. And there are people that if they've worked with very, very senior level people, they've been saved by

Harpreet: [00:05:15] Them and so they'll value

Speaker3: [00:05:16] Those people. But if they've never, you know, it's kind of a self-reinforcing problem. So if they've never been

Harpreet: [00:05:22] On a team with very

Speaker3: [00:05:23] Senior level people, they don't really understand why you need them and what they save you from doing and how they add to the team and make a whole lot of what you produce better and more relevant and how much more credibility they have. I don't know what it is about gray hair, but as soon as I got it, the C-suite started listening to me, not laughing at me.

Harpreet: [00:05:42] And I do not.

Speaker3: [00:05:43] It literally was like overnight I got some gray hair and people started listening to me again. And so there's just so much. You get a lot from people that are in their fifties, sometimes even in their sixties.

Harpreet: [00:05:56] One of the people that taught

Speaker3: [00:05:57] Me engineering like Big E [00:06:00] engineering, mid sixties. And he was amazing at teaching, not only doing but teaching. So there's so many advantages to it.

Harpreet: [00:06:10] But at the same time,

Speaker3: [00:06:11] If you've never had someone like that, you don't understand you get promoted to the level of VP. And now you're just hiring

Harpreet: [00:06:17] People in their late

Speaker3: [00:06:18] 20s, early 30s who all kind of looked the

Harpreet: [00:06:21] Same and all kind of have

Speaker3: [00:06:22] The same level of knowledge and usually find out the hard way that you needed some very, very smart, very, very senior people to augment the younger move fast staff. So, yeah, it's definitely there. But I think we're kind of different in that we're

Harpreet: [00:06:40] Given a little bit

Speaker3: [00:06:41] More of a chance, a little bit longer career path and hopefully it improves. I mean, really, hopefully we can make some some people believers through my sales pitch for people that are older than I am because you may be pushing, I'm pulling for this now. I'm no longer pushing them. So I got to make some space in the field for me in the next four or five years. You know, please don't kick me out.

Harpreet: [00:07:04] I know I'm getting old.

Speaker3: [00:07:05] Please, please don't kick me out.

Harpreet: [00:07:08] Ben, thank you so much for sharing that. I mean, like if if I can be 100 percent real man, like I've got a lot of weight here, but I diet like this beard is completely dyed. It's very patchy in places. I for the first time in so long, I saw like a trimmed off like the the surface layer of the beard and just, you know, how evenly distributed the weight was. And I was like, Alright, this, this looks good, but I'll just wait. I was like, You know, people can start to think I'm too old for this shit. And you know, is. Unsettling feeling. Yeah. Thanks for sharing that. Let's go to to Tom and then

Harpreet: [00:07:45] After Tom,

Harpreet: [00:07:46] If anybody else would like jump in, maybe, you know, want to

Harpreet: [00:07:49] Hear from

Harpreet: [00:07:50] Rashod, Alexandra, Jennifer

Harpreet: [00:07:53] Or

Harpreet: [00:07:54] Anybody else? Please let me know. And by the way, if you do have questions, let me know in the chat, I will add you to the queue. Whether you're watching on [00:08:00] LinkedIn or on

Harpreet: [00:08:01] Youtube or

Harpreet: [00:08:02] Right here in the room, Tom, go for it.

Speaker3: [00:08:06] I remember before my hair turned completely gray, before I even had one gray strand. I got fed up with the way this one woman was treating me and I said just went up to her and said, What the hell? She said, Well, I just have a problem with young people. And I scolded her with my experience, my age, my background. She treated me differently from then on. But now that the hair is completely gray, then I sometimes wonder, am I getting it from the other end if I am? It's very nefarious and stealth, so not sure. It does seem the gray hair I feel like in the experience. Didn't that make you a bit more bulletproof to the bullies out there like LinkedIn police and such? They'll still rear their ugly heads and sometimes they do it in a nice way anyway, and that's appreciated when you make a mistake. But I feel really ignorant on this on the older end. And then I was wondering, you don't look very old even with your slight gray streak, and they're very slight. But I'm wondering if anyone I think I'm the oldest looking guy here just because of my hair color. I know you all think that if I dyed my hair, I'd look younger than all of you. I know you all think that. But no, I'd probably pass for maybe late thirties or early forties. I don't know. I wish I had more to give, though I really have a really feeling little loss on the older end. I just remember experiencing it really hard because I looked young for a long time on the young end.

Harpreet: [00:09:52] Thank you so much. I forgot who exactly I still will go next, but you know, Rashad or Alexander or Jennifer want to chime in, please let me know. And [00:10:00] by the way, I do got mark in for a question on on Data warehouse. So if you and I think there, George hit me up on a on LinkedIn, I have to get back to you there so we can take your question live and direct. Let's go to Rashad and then Alexandra, then Jennifer. And then we'll go on to Mark's question if unless somebody else has anything to add here, then I would love to hear your opinion.

Speaker3: [00:10:23] Yeah, sure. So I guess I'll preface this. I'm 30. I lead a small team, Data science on that guy that was describing. I've interviewed plenty of people who are older than me for Data science roles. So patterns, patterns, I've noticed they most of the time I found that they were transitioning from another field and they weren't necessarily like statisticians before. Let's say they're like in law, they were in the NBA and they're like, You know what? I really like technical work. So that's like, that's a that's a big thing. They definitely bring a lot to the table, and a lot of times I wish I could hire them. Usually, like, let's say, law that gives them that means that they could probably contribute a lot to certain types of problems right off the bat like, say, language problems, NLP like they probably have an intuition about the challenges of text that would make them very useful in that capacity. I've unfortunately found that I, at least in the specific roles that I was hiring for, that I didn't really need like that specific capability. And so with the limited pool that I had, it did end up being other people in their late 20s.

Speaker3: [00:11:34] It was more like a flaw, like a flaw on my part, a flaw on like the overall position. And just like, can I utilize a skill? It was more like, Oh man, I wish I could utilize this like I wish, but then the other person does better and like, say, the case, the verbal case that I give and like, they don't know, they were able to wrap their heads around the real estate problems that we're dealing with better like in a sort of sort of objective, I guess. So I yeah, as far as like older people with [00:12:00] the stats, statistics, background not have the privilege of interviewing very many of those. Maybe if I was at FAANG or something, I would. But yeah, that's mostly what I've observed as far as, like, say, interviewing older people and like, how does it fit in? Yeah, I don't I. There's no I don't have like a strong answer that I'm just telling you what I've observed.

Harpreet: [00:12:24] Thank you very much. Richard, Chateau Tanisha, good to see you again, it's been a while. Alexander, let's hear from you and then we'll go to Jennifer and then we'll jump into Mark's question

Harpreet: [00:12:34] Unless

Harpreet: [00:12:34] Anybody else wants to say anything on the topic. Let me know if you do have a question yourself. Please do let me know wherever you're watching LinkedIn YouTube or right here in the room.

Harpreet: [00:12:44] Now add you to the queue.

Speaker4: [00:12:47] Yeah, I think I might have a little bit of a different perspective on this just coming from the fact that straight from undergrad currently in grad school, in the process of looking for my first Data science job and the most interesting discrepancy that I've found being at this point in my career is that the assumption of how much knowledge I have on paper and job descriptions versus in the interviews seems to be really different in a lot of cases. So just as a quick example, the other day I was in an interview and of course, the job description had a million and one different technical attributes that they were looking for three to five years of experience. You know, the standard things that I've been seeing in the job market. But then when I went into the interview, the expectation of my knowledge base was a lot lower than what was communicated on paper. And that's been really interesting just from, you know, trying to get my foot in the door, trying to be able to to display what I do know at this point in my career and how vastly different that seems to be on paper versus what's communicated to me. So whether that's opportunities for improved interviewing skills, whether that's ageism, I couldn't really [00:14:00] put a direct name on it, but it's definitely been an interesting experience so far.

Harpreet: [00:14:06] Alexandra, thank you so much. Let's go to let's go to Jennifer that actually just noticed the chat about great comments from Russell, so I'd love to hear from Russell on this topic as well. And then we'll get into Mark's question.

Speaker4: [00:14:18] Um, I've seen it from kind of an interesting perspective, both inside the corporation and in this group, as from a profession exceptionally welcoming of a variety of ages, particularly with a welcoming attitude. When you come into this room, people love to communicate, people love to

Harpreet: [00:14:41] Share

Speaker4: [00:14:43] Acknowledges everything to to this group. I see that in a lot of different places, so I see this type of community repeated in many places across age groups. And so that's why my perception as I've transitioned from peer program management and operations role into more of a Data

Harpreet: [00:15:03] Role has been a

Speaker4: [00:15:04] Very positive acceptance of someone like you, said Richard, making that shift from one career to another at a later stage in life. But I'm not going to put a number on it. John, don't you say a word?

Speaker3: [00:15:22] I was just going to have to interrupt you on that one, Jennifer, because you're not that much on your diamond. Tom, I got you b, irrespective of the gray hairs, I think I've got you beat. But I would have to add and kind of confirm some of the comments Jennifer's made there. As far as getting in, I'm just starting in this group. I've been in Tom AIs group as well. Everything that I hear so far is age interdependent. It's strictly knowledge and especially a lot of the work being done remote. What difference does it make? I [00:16:00] see a lot of that, a lot of people's attitudes. I'm taking some courses down at Purdue as well, and nobody can mention the basketball game last night. That's OK. But watching the people that are in that class as

Harpreet: [00:16:16] Well, it's more about

Speaker3: [00:16:18] Being involved. It's more about knowing the subject.

Harpreet: [00:16:22] And I appreciate

Speaker3: [00:16:23] Groups like this as I'm just getting started. My background is

Harpreet: [00:16:26] Engineering, project

Speaker3: [00:16:28] Management and transitioning to data science and enjoying it. It's seems to be a very fun field right now.

Harpreet: [00:16:37] Thanks so much, John and John. This is Jennifer's brother, John.

Speaker3: [00:16:41] Greg, yes, yes. Yes, exactly. Awesome.

Harpreet: [00:16:44] Well, thanks for joining us, John. Super excited to have you here. We'll get to you after I get to wrestle. Russell has some great comments and we'll get to we'll get to you coast to. That's all. Yeah, go for it. Sure.

Speaker5: [00:17:00] A.m. So I've written quite fast and Furious a few comments here, so I'm trying to summarize a few of them. So my first one, which was in response

Harpreet: [00:17:10] To your initial question

Speaker5: [00:17:11] Before Tom answer, was that, you know, it generally depends

Harpreet: [00:17:14] On the specific

Speaker5: [00:17:15] Experience of the individual as to the age, you know, with ages, etc. So if the person that is old has specific experience in the field that overrides the the ageism bias,

Harpreet: [00:17:30] I think, you know, are you struggling to hear me? Yeah. Yeah. Oh, you got it. Yeah, yeah,

Speaker5: [00:17:35] Yeah, okay. And then I went on to say that, you know, I'm in the older category here. You probably don't need to squint too far to see that I've got some gray hairs in my beard. I don't bother dyeing them or whatever. You know, they're just there. They've only been there for the last two years. I swear I had not a single white hair before COVID happened. I've not had those it, but [00:18:00] just coincidentally so other, you know, incidental stress being locked up in home. I don't know what it is, but they've all come in the last two years and one of my other questions was, you know, I generally think that sorry, one of my other appointments was I generally think that my skills and knowledge set balances out my age as I get older and older each year. So I feel generally secure. But I also commented that I come on calls like this and I see people like Ben and Tom who are, you know, also getting on in years. But they seem to have, you know, a lot of a lot of skills and knowledge in the tank that I don't have, and it makes me feel kind of less balanced. But I guess that's kind of imposter syndrome, you know, we've spoken about that on other calls at other times. And what was my most recent comment? Yes, I went on to talk about, you know, criticism and like cancel culture type of thing. I think there's a general philosophy with a lot of people, especially in the younger generations that have grown up with social media being a constant. But it's not been a new source to them that a very easy way to make themselves feel better about themselves is to throw criticism or shade on someone

Harpreet: [00:19:12] Else, regardless of

Speaker5: [00:19:13] Whether it's legitimate or not. You know, you can feel you can vicariously and superficially feel better about yourself by putting someone else down and regardless of

Harpreet: [00:19:23] The legitimacy,

Speaker5: [00:19:24] As I say, you know, very, very much like the immediacy of Twitter, LinkedIn posts, Facebook. So if you just

Harpreet: [00:19:31] Take something really, you know,

Speaker5: [00:19:33] Not very pleasant and you tend to get a thrill from it, and that's one of the worst things about the modern social media sites

Harpreet: [00:19:41] That I see now.

Speaker5: [00:19:42] I think that's covered most. Was there any comments in there I've missed?

Harpreet: [00:19:46] That's good. Thank you so much, Russell. So, Jeanne, I'm going to get to your comment. But Coast did have his hand up just a couple of minutes ago. Coast up. He went to China with the comment. Please do.

Speaker3: [00:19:59] So I mean,

Harpreet: [00:20:00] Yeah, [00:20:00] I

Speaker6: [00:20:01] Can see a lot of the concerns that are being raised and some of the issues like come down to when you're talking about a senior member of a team, the value that senior members have brought to me in the teams I've worked in the past is that wealth of experience then was talking about right now, it may not be the specific experience of of going through this particular technology stack, but I've seen the previous technology stack. I've seen the previous leap and change in technologies I've seen, you know, the previous change in the way business operates. So managing change often gets easier if we start to include experience in our bill for a team, right? So my question and this is more of a question extension than anything is how much does this come back to the fact that a lot of the data science when you start off in data science, you typically start off as an individual right? You start off as an independent person looking at a bit of a data set and one single script to create something right. Whereas in more traditional engineering fields, you're very aware from day one that your specific skill set isn't going to get you over the finish line. And it's going to be a team effort, even within a group with the same skill set. Like so in robotics, there was never a single project that was actually serious at all that I could even imagine that I could do by myself. Right? But what was he in Data science? Is this. I don't know if it's a biased or if it's a it's definitely a mental thing where we think, OK, I'm just operating on my own. And we've seen that in effect because the amount of time that it's taken to get Data scientists on board to say, OK, we should be using it, we should be using, you know, PhD workflows so that we can collaborate. So the collaborative side of it is lacking, I think, in data science compared [00:22:00] to all the

Harpreet: [00:22:00] Principles of engineering. But.

Speaker6: [00:22:03] So my question is how much does that contribute to this idea that why do I need someone more senior?

Harpreet: [00:22:09] I'm able to do

Speaker6: [00:22:10] This task myself. You know, we're maybe not able to see the bigger picture when it comes to engineering larger and longer living systems. And I'm really curious to know what some of the others who have been through those engineering cycles think of that.

Harpreet: [00:22:25] The great question, and I like this quote that A. dropped right here into the into the chat,

Harpreet: [00:22:31] I think it's appropriate

Harpreet: [00:22:33] To say

Harpreet: [00:22:33] Here

Harpreet: [00:22:35] One of my most favorite sayings concerning intellect and wisdom is from Einstein, and it says an intelligent person

Harpreet: [00:22:42] Can solve a

Harpreet: [00:22:42] Problem that a wise person will avoid. So I think that's a that fits in quite quite nicely here. But while you guys noodle on that, let's go to Gina. And if anybody has any comments to to Kosovo's question, please let me know. And Margaret, I promise we will get to your question. We'll get to

Speaker4: [00:23:02] It. Yeah, and so, Mark, I'm sorry that I missed part of your answer because I think it might have been getting to part of my question, which is, you know, if I think I'm hearing that, you know, the experience that comes from a worker who's been out there for a while and that would be me as well, like that is valuable in so many ways. But you know, if they if they haven't worked in that technical area. So I'm career

Harpreet: [00:23:36] Switching into Data

Speaker4: [00:23:38] Science, although I've done a lot of analytical work, I'm done consulting, I've done a lot of things and I've, you know, you keep hearing, you know, kind of referring back to customs comment. I hope I said your name, right? Cos you know that maybe Data scientists aren't as collaborative as they could be. And yet you hear over and over again, [00:24:00] Data science is a team sport. It's extremely important to be able to collaborate across different, both within a data science team and to different business units. People with more experience tend to, you know, not always but tend to be able to navigate these things and see patterns from the, you know, past experience that they say, I've encountered a situation like this before, and I think we need to proceed in x y z way. And yet if you're hiring for somebody and you're looking first and foremost for certain technical experiences, you know, then that. Whether it's soft skills or just other kinds of experience that are

Harpreet: [00:24:51] Relevant but not

Speaker4: [00:24:52] Necessarily easily communicated in response to a job application or in a resume review, how do we how how do you communicate your value in terms of that, those years of experience and how you can bring that to a position? Is it that you should be targeting manager positions or are manager positions really only feasible if you've already mastered at least certain parts of the tech stack? So I guess that's my it's my one question to the following question. So yeah, maybe you want to come back to it and people can noodle on that after responding to customer?

Harpreet: [00:25:42] Yeah, definitely. Well, how about we do this to guys noodle on both these questions because we're really good. But I'll have Mark and Jeremy provide some insight here to this question. And then right after that, Mark will go right into your question. But this is not to say that we're tabling the earlier discussions. If you guys [00:26:00] have some some insight definition here, so please do share it. But. Mark, Mark and Jeremy go for it.

Speaker3: [00:26:06] I hate to ask Moz the core question because I turned my computer off and came back, and I know we're talking about age. I've been trying to figure out exactly what the real question was.

Harpreet: [00:26:16] Yeah, so it was a combination of questions, so do you want to kind of quickly summarize the one liner version of your question? Then we'll have Mark and Jeremy. Give some insight on that.

Speaker4: [00:26:30] Sure. So and Costa may want to as well. So basically for older workers, people may be transitioning old or not. How do you what's the best way or what are some ways that are effective in terms of communicating the value of your experience? Because a lot of times your resume will be shorter and because you're only focusing on your data science experience, people might not realize what all you bring to the table. I think it's a tough problem. I mean, obviously. But yeah, I would love to know your guys's thoughts on that.

Speaker3: [00:27:08] I want to go No. One personally, I come from the finance background and I've been to Data science thanks because of the problems I've been facing, so every time I was looking for answers, I didn't necessarily want it to have like technical information, but more like experience on how to deal with things. And I think this is the big, big point on. It's not about like coaching or telling the newcomers how to do things. It's more about like how to mentor them and how to give them a way of thinking based on the expense. This is what I was looking when I was talking to older people, and I've been specific on

Harpreet: [00:27:49] This when I was talking to

Speaker3: [00:27:51] Them and they were happy enough and wise enough also to not be like, you should start with this and this because their technology was. Was not what I was [00:28:00] looking for, I was looking for Python, I was looking for a bunch of notebooks that I could start from, and those guys were using VBA and low fashion technology that were that was working before. So I guess it's more about the pathway, how to think of the problem, how to isolate different hypotheses. And I learned a lot from the from. I've been mentored by a lot of like have been looking for mentors every time I was trying to fix a Data problem in my past. And I think this is where we as a younger generation, what we are looking for is more mentorship than than coaching or and stuff like that. So mentoring is is probably my answer to this question. It's about mentoring. And then my take for things like, first of all, I really value different perspectives, I come from startups, which are typically on the younger side. And and so I'm trying to think back. I don't think I've had many managers or colleagues too much older than me. So I'm now reflecting on that. I'm like, What do I need to do to change that? But the second thing is I constantly reach out different perspectives, especially people who have substantially more experience in the industry. A big reason why I come here because individuals who have industry experience are here and so I can hear best practices and bring it back to my to my company. But to go back to your original question of, you know, you have these amazing skills and you're going into data science and you are complementary, so how can you bring that to the table? And for me, at the end of the

Harpreet: [00:29:44] Day is like my job

Speaker3: [00:29:46] Isn't to mess with Data or my job is to write code. My job is to provide value and just so happened to use Data as my medium to do such. And so how can your previous skills before Data [00:30:00] really allow you to drive impact? So for me, like I talk a lot about my adventures and entrepreneurship, and I'd say all the

Harpreet: [00:30:06] Time, like

Speaker3: [00:30:07] Being a data scientist did not make me a better entrepreneur. But entrepreneurship entrepreneurship made me way better data scientists. And so I know how to take like a really raw idea and quote unquote bring it to market within my company and get buy in. Who's going to need it? Who's my core user and really approach this product focus and customer focus where my products that I build internally are accepted and proliferate throughout the company? And that's not data science skills at all. That's the fact that I've gone through that we're trying to bring my own kind of company product to market as replicated as smaller scale within the company. And so, you know, how can you tie those previous skills to show that I can take this Data medium and do the same thing for you? That's my huge value. Add. Oh, and by the way, I can code really well to.

Harpreet: [00:31:01] A great comment coming here on YouTube from Dave Spleens says that older workers with the experience should be held in high regard, as many systems are working with legacy Data and legacy systems. F AIs in Milwaukee still uses and trains in Fortran. Interesting. So, yeah, let's go to Mark's question. By the way, if anybody if you're watching on LinkedIn, if you watch on YouTube, if you're watching on here in the room, you got a question. Let me know I will ask you to the queue. There's a lot of stuff going on in the chat here, so you might have to like just send me the DM so it stands out. And if anybody wants to come back to any of the questions that was posed by Kosta or or Gina or Coastal Virginia, if you're going to resurface those questions a bit later. Please feel welcome to do so. But Marc, let us go to your question.

Speaker3: [00:31:54] Yes, I'm working on a really exciting project at my job, where essentially [00:32:00] it's scale or Data warehouse make it. So it's a really strong foundation to really build on top of our future analytics. So we have a data warehouse right now. We're a startup we're scaling and now that we're scaling, we're coming across kind of growing pains, right? And so I've been tasked with leading a team to to explore where options are and then afterwards build it. And so currently, I'm in that exploration phase and I want I don't want this conversation to be like, what technology to use, but like what pain points or things to consider. If you were to build a data warehouse to like catapult your data strategy

Harpreet: [00:32:40] And

Speaker3: [00:32:41] Your your your data maturity, like what things would you consider like must haves or a avoids?

Harpreet: [00:32:48] Let's go straight to Vin on this one, and after Vin, is Mexico still here, if Mexico is still here, I'd love to hear from you. And then maybe even if Roshad or Alexander or anybody else has some tips here would love to hear it. Just, you know, just give me like that. Hands up reaction thing. And I can actually think you go for it.

Speaker3: [00:33:10] Yeah. Was going to kill it, but in advance of her absolutely destroying this question, making it sound simple. Why are you doing this? So I think I think a big why is at the end of the day, you know, I think I think the best way I can say I'm not trying to put like information out there for a company, but you know, we're building fast trying to find product market fit. We found that. So now we're scaling. And so now we're like, All right, we we're able to for the current level we're at, we're able to deliver pretty well. But to get to the next level, our data infrastructure is just not cutting it. And so we need to invest and make upgrades to it. So that's why we're engaging in it is there's these different opportunities in the market that require data processing. And to get [00:34:00] there, we need to do this. Well, let me ask the vague questions, because I kind of know what I think, I know what you're doing, but is it the amount of Data or is it the amount of data that you have to send to or the amount of models being sent to? Is it all three or is it one more than another? I would say, I think the best way I can describe it, I described before our Data structured really well for a web app, not for analytics.

Speaker3: [00:34:30] Our engineering team crushed it and our web app Amazing, Super Secure can deliver what's want what it is like. Awesome product. But then that's dumped into the data warehouse and I've just been running around trying to make sense of it and put some like stop gates in there to get it running. But now I've gotten because I've done that gone by in from leadership in the company of using these tools like, Oh wow, this integral, how can we go back? And so my amazing man just jumped in. Well, we can just upgrade it and make it much better for you. Yeah. I mean, it sounds like transformation like Data transformations, like the number one thing that you need to do, and it's not so much cleaning as getting it ready for to make it useful for models to make it useful for. Mm hmm. For ingestion and for inference or for training. Are you like, are you targeting both? Yeah. So I think I think the best way to describe it is one the Data pipeline from the database to the data warehouse just making that very consistent. And then the next step is, all right. We got this raw data. I'm pushing for DVT because I love that transformation tool, but I'm open to other suggestions and the sake of being exploring what's the best solution for our use case.

Speaker3: [00:35:47] But there's that transformation piece basically create like a really solid Data mark. Once we have that data mark, we can do the cool analytics. We can create different features and start building machine learning models and then serialized, then put it back into the product. [00:36:00] So at the at the core of agony, I strongly believe that data warehouses really serve as the foundation for driving Data maturity. And so I have this opportunity to get it right, and I want to make that happen to make my life easier in the next few years. It's the reason why I've been asking all these boring questions. Now you have to do what I just did with you to everyone else in the business, because you're going to find out that everyone has a different opinion. And what you put in place the entire company and I know it's small. That's why I think this is actually feasible to do by one person. But whatever you put in place, you will either be loved or hated for. And if you don't understand how everyone else cares about this because they will and the tool that you're putting in place, you'll only find out after it starts messing with their

Harpreet: [00:36:51] Stuff that they cared.

Speaker3: [00:36:54] And then they're going to be like, Well, why didn't you ask me? Because, because how was I supposed to know? So that's this is what I've saved you from is now have this conversation with. Start with the why? Why are we doing this again? What's your OK? So what are your use cases like? Seriously, just hit exactly what I did with everyone who has ever

Harpreet: [00:37:12] Looked at Data in the entire company.

Speaker3: [00:37:15] So a couple of things. One, my manager's phenomenal. And she already organized a whole team and just said, Mark, you're leading it.

Harpreet: [00:37:23] So again, my

Speaker3: [00:37:24] Manager is amazing. So we got this cross-functional thing and I did that roadshow of asking about Data a year ago and asked everyone to have a whole spreadsheet of that and of the whys. And that's how I got to the building. Like these, like the one tools to make their lives easier. And now that's why they're like, oh, data warehouses are important. We should invest in this. So I guess, like from there, you know, I've done that groundwork will be worthwhile. I'll do it again since, like a year later. And it's kind of like being more serious or, you know, is there another step after that? Now you've forgotten somebody. Just trust me. [00:38:00] I know you've gone. I know you've gone through the entire organization, and I know your manager

Harpreet: [00:38:04] Is a rock star. But one hundred

Speaker3: [00:38:06] Percent, you've forgotten somebody and it's doubled since then. So I probably. Yeah, there's always. There's always. And so, yeah, go through, do do the questioning just kind of like I did and figure out what the biggest concerns are and try to get to the point where you consolidate to as few tools as possible, building as few tools as possible, you know, buying as much as you can off the shelf and make it as easy to use and easy to train on.

Harpreet: [00:38:36] Because as your business scales, you have to onboard

Speaker3: [00:38:39] So many different functions. And since you're a

Harpreet: [00:38:42] Startup, you're going to be onboarding

Speaker3: [00:38:43] Like crazy. When you said the word growth, the reason why I'm telling you to do the roadshow first is because you have to know not only everything you know now, but everything. You don't know that you're going to be using it for. You are also going to need to know what the people you haven't hired yet might need. And I mean, that's really where you're going. And that's why I'm kind of leaning on the business side of this so

Harpreet: [00:39:06] Heavy is because

Speaker3: [00:39:07] You want to build as little as possible because it gets expensive and ugly. If you have to build a lot, you want to pull as many off the shelf tools as you can, because that'll make training time easier.

Harpreet: [00:39:17] It'll be easier

Speaker3: [00:39:17] To hire skills and capabilities going forward. If you have something that's common and off the shelf not customized, and you're also more likely than not going to be able to

Harpreet: [00:39:26] Scale using off

Speaker3: [00:39:28] The shelf to new use cases. Because I mean, every company that's selling this, every large company that sells any sort of data warehousing or that sort of solution has like an entire ecosystem. And the bigger the ecosystem is, the more likely as you go in and niche use cases because health care has them that you're going to be able to actually find a solution provided by and compatible with what you've already built. And so those are like the businesses and nominal let Mexico be technical and kill it.

Harpreet: [00:39:57] Go to Mexico, and then Russell's got some great [00:40:00] tips in as well. So after Mexico go to wrestle and if anybody else has questions, please let me know in the chat I see y'all and LinkedIn Kennedy were watching the other smash at Lake. And yeah. Let's go to Mexico.

Speaker4: [00:40:15] Oh, rollout and maintenance, those are really, really, really big factors because whatever you build, your team will be owning, which means you will be the one who is waking up at 11 p.m. to those like Slack notifications from Stack Driver, that something has broken and you're going to have to handhold people as they use a new tool and all this so. So we're going through something. Well, not necessarily for storage, but we're setting up infrastructure on our team and we try to think of what we need in store like three phases. The first one is like what is immediately relevant to the project. So it's like the short term phase. The second phase is like the rollout, the education, the you know, the drumbeat messaging you do at the town halls, the office hours where you lead people through it and then what we call the longer term success, which is the maintenance and ownership of it. And because what sometimes happens is like when there's this kind of white or gray space and you're you go into that vacuum and you take on leadership and you build something that becomes the thing that your team owns forever and ever and ever, and they never get to touch anything else like, that's just what happens. So. So a couple of things that I would think about. One, the long term maintenance, that's a huge thing. It's easier to maintain something if it's within the same vendor and especially if it's a managed vendor in general. So even though I really, really love the open source out there and I love the small, gritty startups, the reality is a lot of times like your existing provider like Azure [00:42:00] GCP, they're more than capable of providing robust technical training

Harpreet: [00:42:09] And support,

Speaker4: [00:42:10] And more importantly, to you can negotiate with them for credits or for like like consulting. That's my favorite thing is I'm like, Well, you're a big company, you're Google. You're big. Let's negotiate. Let's let's figure out some credits. Let's let's squeeze you a little bit because it's not even a squeeze. It's like a it's like a tap on the side of a whale. They're not going to feel it, you know? So that was preferred. I think it's good to be really boring and unsophisticated with the tooling, because if it is really easy, then the fun part is later on, you can convince some other team to like, take it on if you no longer kind of want to do it. So that's important. The second part is security. When you have to roll your own stuff, you really have to think about security and credentials. So, for example, if someone leaves and their model is still running in production, but it's like a personal account. I've seen this happen before, but it wasn't like a service account with credentials that you could manage. Then that's like, not good. So security is a big issue. The other part is also like connectivity and just also how fast off the ground you can kind of get going, you know, so those are things I would I would think about. And in the world, right, the really sort of like unsexy kind of solution that just happens to work for most people is like BigQuery with like depending on what kind of asset they need with, like a looker dashboard. It's I mean, looker sucks. But sorry for anyone who's like a looker fan, but you know, in comparison to others could be better.

Speaker3: [00:43:52] I use Data studio, and every time I use that one a screen,

Speaker4: [00:43:56] Oh yeah, I think didn't I thought someone said that they are sunsetting Data studio [00:44:00] and like a couple of years or something? I don't know. I should. Yeah, they probably should. But the way so the way we do it is we use air flow to manage like essentially these like Python script

Harpreet: [00:44:11] Pipelines that

Speaker4: [00:44:13] Pull like Data to BigQuery. If we have like stronger use cases, then we can use something like spanner a big table. And then we have some hub sub or Data flow throughout that. So but it's it's like really boring, right? Like, you don't see anyone writing white papers. The more boring the solution is, the less often you'll see blog posts about it, which means that like all the companies that are doing something that actually works and is like scalable and like, robust. They're not writing about everyone who writes the blog post in the case studies and system design reviews, it's all like systems that either they had to like, finagle something to make it work or like, I don't know, they're using some like buzzwords stuff, but a lot of like the typical sort of like use cases for batch and streaming. They're pretty well documented within like GCP and Azure. And so that's another big selling point to why it's good to like, go with those. But I would just like one hundred percent lean on the consultants and and like the solutions architects from the cloud companies and like, basically, yeah, like that's that's the way. Like, I would do it. Yeah, just be like super vanilla and all that, because the thing that you still need to do right is like. Just getting the project up and running, that's like the first stage, getting people using it and adopting it's the second stage making sure that it doesn't like kill your mental health and your physical health with like pings late at night. That is like the tail end, but is like a very important part. So I would definitely think about it in those three phases.

Harpreet: [00:45:57] Yeah.

Speaker4: [00:45:57] And in terms of like the actual solution, like [00:46:00] if, for example, if you have something super specific, like IoT streaming. They have like these cases, they're in all those vendors I've seen. So that's pretty well documented. But even then, like. Yeah, that's yeah, that's just that's just my take. Use very boring tools,

Harpreet: [00:46:24] Kabukicho, let's go to a wrestle, then I have to wrestle. We will go to Rishard then Kosta. I had cued up or shot there.

Speaker3: [00:46:34] And also, real quick, this for contact, I think the best way to describe it is that I'm not implementing a new data warehouse. I think the best way I can describe it is refactoring our current data

Harpreet: [00:46:43] Warehouse to make it better.

Speaker4: [00:46:49] Even better to to squeeze as much as you can out of the cloud vendors, just squeeze them. Get that insight. Get that strategy.

Harpreet: [00:47:01] Russell, go for it.

Speaker5: [00:47:02] Sure. My comment was about the vendors themselves and probably more in context if you are looking to go new mark rather than reflect what you got already. But it was an understanding that many might not have that with a with a large scale enterprise data warehouse. They will store the data that you've got in places and they'll have a data center infrastructure beneath your interface to

Harpreet: [00:47:28] Them that that stores that

Speaker5: [00:47:31] And most likely to maintain diversity and security of the data. They will store it in multiple locations. So for a country that's as big as you know, the North America's, those multiple locations may be within the country. So the only complications you've got for, say, GDPR or equitable data protection regulations or laws will be across state. However, if they're going like a next level, and I suppose this is going to depend

Harpreet: [00:47:58] On whatever the

Speaker5: [00:47:59] Company contract [00:48:00] is with the provider, if they are duplicating that data internationally to protect against, I don't know, cataclysmic events like meteor hitting or something wiping out a huge amount of data centers which which they may choose to do. As I say, depending on the on the contract you have

Harpreet: [00:48:19] With them, then you need

Speaker5: [00:48:21] To be mindful that if you have specific data protection regulations applicable to the data from a client's data or your own data, be mindful that there could be some

Harpreet: [00:48:34] What I call

Speaker5: [00:48:35] Kind of stealth issues with where that data is stored. So a lot of people don't think of that layer beneath, you know, the data warehouse itself, but it can be an issue.

Harpreet: [00:48:49] Go to Rashaad and then post.

Speaker3: [00:48:55] And it was very interesting, I learned a lot from everyone else's response, I don't know as much about the technical side of this, and so that was very cool. Thanks everyone for that. I just want to add like there is some small things. So first, there's a I heard a lot of talking about like what you want to do and then you generally want to serve people more and that you made this list of interviewing people from a year ago, right? So to I learn from a chief product officer I served under once. It's a good tool. I think all product managers

Harpreet: [00:49:26] Use is to figure out

Speaker3: [00:49:27] Like your customer like archetypes like who are the different populations and the company and then like abstract them as much as you can and then figure out. And then from that, you could be like, who's being the least served right now? It sounds like analysts or people who want to do more deep. I mean, it could be data scientists and model training, but it sounds like if you're talking about this, you're probably first wanting to, like, just understand the data in the first place, like visual make, make like dashboards like sort of thing like, understand you, it's the employees serving our customers. So [00:50:00] customer success? Oh, customer, OK, yeah, because I can work with the data, it takes me a little bit, but I can easily get the data to do analysis. But like, that's only a few people on our team. We need the whole company who aren't data literate or not data illiterate, but aren't code literate right to be able to say like, Oh yeah, this is our data and go to my customer, be well prepared. Mm hmm. Mm hmm. Mm hmm. I read earlier today someone was opining on the death of the the star schema and like the dimensional model just saying they were saying that in the eighties, this model is sort of created because it allows you to efficiently store data in a very highly normalized

Harpreet: [00:50:42] Form, and you could have indices

Speaker3: [00:50:44] That would quickly let you do whatever analytics you want. But it's not very intuitive for people outside of Data to understand that model and then do the joins or whatever to make it work. And they're like, Well, people want to see spreadsheet and spreadsheet esque data, and that's like more broadly understandable outside. So my suggestion is like, OK, if you've identified that population, then start with like work backwards from what ideal state they would see like on their end as a user. And then you can be like, All right, that's the format of the data, because it really sounds like you're just reconfiguring the form of the data to be like more wide, more long or whatever. Yeah, exactly. No, no SQL database. So it's very nested in BigQuery allows nesting of data. And so when I first showed up, I it was pretty scary. With no SQL databases now. Oh, scraping JSON, I understand. Yeah. The other thing I'd say, like just for tools like if you're thinking of tools, don't just look like who has the most features. Also, consider who has the most momentum and then who will be the best partner in the long run because sometimes like products will, they'll grow into what you want and sometimes depending on how big a customer you are, you can influence them and help make them do customization [00:52:00] stuff for you. I think like also, oftentimes if a product does a lot of momentum but is not the leader right now, you'll get a price discount, and I think that's important in startups. So if you grow together with them, it's it's it's a call, right? You don't know, but I think they're probably a startup would especially benefit from looking at the momentum of when comparing and not just like listing the features or being like, All right, this is the best. It's like that.

Harpreet: [00:52:31] The coast of Mexico, and then anybody else has questions or anything, please do let me know.

Speaker6: [00:52:38] Yeah, look, I mean, the two things that come to my mind in this conversation is more how much support can they actually provide and how how serviceable, how does that service actually work, you know? Are they large enough to be able to support high volume Data flow? You might have a system that's full of brand new features that are, you know, cutting edge. But if there's still a fledgling startup or a fledgling

Harpreet: [00:53:05] Business, they might not

Speaker6: [00:53:06] Have the support capability to see you through volume, right? So you've got to make that phone call between momentums visibility, their ability to support issues, right? And this is where some of the bigger players

Harpreet: [00:53:20] Win out is

Speaker6: [00:53:21] That they're able to provide that level of support if you need it. But yeah, it's a bit of a balancing act. And I found myself trying to figure that out myself and various other peoples, not really for data warehousing or Data legging, but more on the one of the Data labeling and that side of life. Yeah.

Harpreet: [00:53:43] Thank you very much. Let's go to Mexico.

Speaker4: [00:53:47] Yeah, I guess, like it's I guess in my head, like I'm I'm still in some ways, I'm actually not hearing a need for a data warehouse as opposed to like. Some sort [00:54:00] of analytical. Tool slash, pipeline analytics, engineering like layer, I guess. Yeah, because I'm like, well, if it's all ready in BigQuery, I have like some notes on like the comparison of like latency and query calls against BigQuery and others. So I'll just see if I can try to find it. But yeah, because if they're only doing like analytical queries and some of them like are not or are not familiar with SQL. It almost just seems like you kind of just need some interface on top of the Data storage, like on top of BigQuery understanding, like are there ways to sort of like better kind of like shard or index BigQuery for like faster optimizations? And then for people, if you have like two specific sort of groups, then it's like figuring out how you can enable one group to use SQL on the other one to just do like export to CSV, which in my experience, like a lot of the customer success revenue people I work with, that's what they really like. They're like, we either want it in Salesforce or we want it in HubSpot, or we

Harpreet: [00:55:25] Want it so that we can like

Speaker4: [00:55:27] Copy it to Google Docs and like do charts or whatever. So it's almost like, I don't know how much you need to build as opposed to like just plug it in or what have you.

Speaker3: [00:55:43] Yeah, I mean, I think it really called it out. Is this an engineering piece and and been called out to the transformation component is that once we get those transformations and it's stored in a data warehouse in a way that's easy to access, I can easily plug it to Looker or [00:56:00] Data studio today. I can easily plug it into like a notebook system and do R&D for like machine learning and stuff like that. And I think my hypothesis is that a big bottleneck for our Data maturity is the fact that we don't have this transformation layer within the data warehouse. And I feel like once we get that key piece, then I can start pushing for other aspects that are more fun, right? Doing the cool analytics, doing the machine learning. But none of that stuff matters. If every single time I do it, I have to spend like 20 30 hours preparing the data because there's so much business logic and so much, so much going on or the data pipelines down for certain tables. So there's a mismatch in tables. So my my hypothesis is if I get the data warehouse like a solid state, then we can move forward on other things. But we skip the data warehouse. We're just putting Band-Aids on top of things on top of the real issue, and we're going to have a shaky foundation for when we want to do machine learning, when we want to do advanced analytics, when we want to do a b testing in a much more advanced way.

Speaker4: [00:57:06] So actually, part of the pitch might be as part of the pitch to build out the team to like get the people because essentially like I feel like in most cases someone has to be responsible for doing the transformational layers. Sometimes it's like a specific team. And other times it's like part of the duties of a Data engineering team. It almost seems like you might want to push the head count aspect of it while you're figuring out the technology solution as well.

Speaker3: [00:57:34] We closed a Series C, so

Harpreet: [00:57:38] We we

Speaker3: [00:57:38] Have funding for

Harpreet: [00:57:40] For some

Speaker3: [00:57:41] Roles and I'm excited to announce some to to this group once they're there ready for that. But also like, yeah, we don't have a data engineer and like, we need a data

Harpreet: [00:57:51] Engineer

Speaker3: [00:57:53] And I've just been filling in that spot and I might just try to move towards that Data engineer piece because it's [00:58:00] just a big need in our company. One of the most arbitrary. Don't think. Don't you think you or AIs should be marched next CDF So.

Harpreet: [00:58:10] She was chosen for it, she Data science officer.

Speaker4: [00:58:17] So here's a go to play the devil,

Speaker3: [00:58:20] I'm just dreaming, Mark, just dreaming. Sorry, John has that title, unfortunately.

Speaker4: [00:58:26] So to play the devil's advocate, if he maybe actually now's the right time to find the right Data engineer, because the messier things are the more pain people feel. Whereas if you solve it, then they'll be like, why do we need to find someone?

Speaker3: [00:58:39] So just a thought. So just a further question on that and please stop me. I feel like I'm taken up a lot of airtime is what team should a Data Data engineer be on? Should they be under the Data science team? Or should they be under the software engineering team because you can make arguments for both? But typically our thought is we think it should be on the software engineering side because they're the one creating the platform to like, collect the data and that transformation from them to us. It's very challenging. We don't speak the same language, so we had someone on their side that can speak our language. That could potentially be helpful, but it's just guesses right now.

Harpreet: [00:59:26] So I want to go to this one because

Speaker3: [00:59:28] I'm curious as we

Harpreet: [00:59:30] Responding very positively to this. But I will say, if you tune in to the episode of release today with Alyssa Simpson Ralph, where we talked about this very issue like, OK, who do you hire in what order for a data science team and

Harpreet: [00:59:44] Check that out?

Harpreet: [00:59:45] Then let's hear from you.

Speaker3: [00:59:49] You will hate your life if you don't control everything that you need in order to deploy a model to production.

Harpreet: [00:59:54] You will hate your

Speaker3: [00:59:54] Life if you don't control everything you need to deploy a model to production. You will absolutely hate your life, so [01:00:00] everything that everything that touches a model must be controlled by the Data science

Harpreet: [01:00:08] Team, which is awkward

Speaker3: [01:00:10] Because you're going to be building a data and analytics organization. And this is going to start touching other people's domains. So this gets awkward quick.

Harpreet: [01:00:21] But if you can get

Speaker3: [01:00:22] In front of it, which is where you are right

Harpreet: [01:00:24] Now, where they haven't

Speaker3: [01:00:26] Been hired in yet, and so you don't have to start stealing from other organizations yet, bring them all in under and start like a data and analytics organization like start using that word or those words, because that's what's going to help senior leadership

Harpreet: [01:00:42] Understand. It's not just a

Speaker3: [01:00:44] Data scientist, it's not just an ML engineer. It's not just

Harpreet: [01:00:47] Like this little wing

Speaker3: [01:00:48] Of it now that your business basically depends on. Data science and depends on models, the Data science organization is part of the data and analytics

Harpreet: [01:01:00] Organization and own

Speaker3: [01:01:02] Everything from infrastructure to people that you need to deploy models because if you don't, you will have people in charge of part of your workflow or part of your development and deployment pipeline who don't know what you're doing. And that will always be a process of painful education. So Data

Harpreet: [01:01:21] Engineers, machine learning

Speaker3: [01:01:23] Engineers, even some software

Harpreet: [01:01:25] Developers are going to end up working

Speaker3: [01:01:27] In your organization and start to consolidate just from the beginning. That's super interesting. Ok, I'll bring that to my manager because we were having this conversation, just trying to go back and forth, like, where did engineers sit? And so we're kind of been on the fence on that. It's really interesting.

Harpreet: [01:01:46] And I mean, just intuitively, I would I would imagine it should sit. They should that should sit on the data science team. That's how I would would structure it as well. But let's hear from Rashad.

Harpreet: [01:01:58] Yeah, so I'm [01:02:00]

Speaker3: [01:02:00] I'm with Ben on that, so we recently we were pulling data from a very gifted Data engineer we have and we were pulling data to do machine learning on from a view in SQL. And The View was like three years of data. And everyone, everyone was under the impression that this was all the data that existed. You can see where this is going. The new year happens and the data vanishes from twenty eighteen or like, Oh, well, that's interesting, what happened there? And so we're like, OK. And then we went to the we asked a bunch of people and eventually found someone who could point us to the underlying query. We regenerated the data and we found it went back all the way to twenty thirteen transformational. It's like, Oh, interesting. I'll always ask tough questions now, no matter what, because we were trying to go fast and we're like, Oh, we didn't ask that question. So, yeah, you want to control it, you want to control it. I also think in a broader perspective, this gets to the constant debate between functional teams versus pods. So like a pod is like sort of oriented long, a long term team oriented with like made up of several functional pieces that are doing a specific thing. Yeah. And and it sounds like you guys are functional teams in enterprises they often do. They call it the matrix like the Matrix. Org, like functional teams arranged into pods over, you know, and that can get complicated. I i it probably depends on your on where you are and your use case, but I would

Harpreet: [01:03:36] Generally think I

Speaker3: [01:03:38] Generally think like the pod orientation makes more sense because the team gets used to working together. And then there's always a business connection that is implied in the team structure rather than a functional team where it's like, I guess we should be doing stuff. We should be doing engineer stuff. Let's build some pipelines, I guess. You know, meanwhile, like the pod, it's like all the businesses in the business. So that's [01:04:00] my hot take. Vicki, go go for it.

Speaker4: [01:04:05] Yeah, I would say, like it's interesting, though, because like typically. Tech companies think they're doing Data products, but a lot of times they're they're not like. So it's one of these things where they have like attritional like I.T., diva or Google system, where everyone suffers. No one's happy. Even the Depay's themselves, right? But they're not really doing Data products. And then like you build a data science function, they're like, Oh, we need data to build like machine learning and data products, right? And then you usually have like. Though, you know, the the old guard where they're like, yeah, no, we just can't do that, we can't trust you. Yada yada yada yada. So it's a little bit weird where it's like, sometimes you have to do what they call like a skunkworks skunk works team. Is that right? Or like, you kind of have to put together this like lean Swiss Army Ninja team where they kind of go do. Stuff that maybe is not cutting edge for the industry, but it's like cutting edge for the company. They have to choose their tooling and they're essentially proving out the concept of like, Hey, actually, we need to be much more seriously invested in producing these Data assets, making sure that they work.

Speaker4: [01:05:22] Sometimes even using new tooling. And then the rest of the company will kind of catch up and then they'll go like, OK, well, now we need a Data engineering function. So it's one of these things where it's like just because you start the first hire like in the data science team, it doesn't mean that that is where all the efforts will be concentrated and. And to a certain extent, you wouldn't want that long term, right, like you would kind of want a Data engineering or Data function that does sort of touch more teams, that provides value for the company. But sometimes you have to start off like a really small place or like in a specific team [01:06:00] to show the rest of the company that it's worth like heavily investing like in a Data function. So it's one of those things where they'll start off in Data science and then once like enough value has been proven. They'll say like, Oh, well, we need this for like this org like finance and we need it for customer success and we need it for marketing and all that. And then they'll be like, Oh, but we can't just put it all in this one person and then they'll go like, OK, well, we know how valuable this is.

Speaker4: [01:06:23] So now let's build out a function and then it'll kind of go towards there. But like, if you don't start that person out in the high value area, you're never going to be able to like, prove out the need for it. And this is also why sometimes like instead of trying to like, solve someone's problems or solve a company's problems like, OK, I'm going to go in there and I'm going to figure out a technical solution, sometimes it's good to kind of like, let them suffer and go, Yeah, see if we had more people we could, if we had more people who are strategy minded who could come in and help determine the stack, you know, then we could, then we could kind of go further and all that. You know, it's one of these things where when things are going well, companies are like, you know, and things might not be going well, but you have like that one super service oriented person that is like working 40 to 60 hours a week, they're covering up all the patches. So the company never sees like, oh, like, this is a pain point.

Speaker3: [01:07:24] So my gosh, I'm not.

Speaker4: [01:07:28] I'm not targeting you, mark. That was really like my last three jobs, you know? But so it's one of these things where it's almost better to like, have the problem, be there like like make the argument to hire a person into the team, like hire a data entry or analytics engineer into the data science team, have them like, prove out some really cool projects and then also have them like get buy in, like help determine the stack. And if you do that, you can kind of make the argument for a more senior person like, Hey, we would want someone who's [01:08:00] a senior engineer or like a staff engineer or whatever, because they'll help determine the stack. They'll eventually help pave the way to greater value. And then once they've done that right, you have the skunkworks team. That's awesome. Then people are like, Oh, well, we want that too. It was two of those. It's kind of like how shopping right there. Like, I didn't want a house, but I want that one. It has a pool. It's got a fireplace. It's really nice and cozy. So that's kind of like probably how that's like the life cycle in a way, but for sure, like starting it off into data science, I think is really important for that Data entering function and also making sure that they're able to play nice because there is a little bit of this territory, sort of like Big Dog kind of fights that go on where it's like, Well, we own the database and we can't trust you. So if you get a solid engineer who can talk the talk, then they will be more like amenable

Speaker3: [01:08:56] To that literally took me a year and a half to get that bi engine. Give me access to the database. Yeah, yeah.

Harpreet: [01:09:06] Great insights. Thank you very much. Go. Anybody else got any questions or anything to add to this discussion? Please do. Let me know. Otherwise, I'll start getting some PSA public service announcements. Great stuff happening on the podcast coming up for the next few weeks. Next week in episode releasing with Liz Gosselin, coauthor of No Hard Feelings, one half of Liz in Mali, who is also at who and who Marc introduced us to. So Mark, thank you so much for making that happen. That was a great conversation.

Harpreet: [01:09:38] Liz is amazing.

Harpreet: [01:09:40] We recorded that podcast a very long time ago, but I'm excited to be releasing that finally. Then after that, we got a episode with Justin Winn, who is if you guys don't know Justin Lin, follow him. He's a he's pretty much doing some awesome stuff gearing content on LinkedIn towards college students, which I think is really [01:10:00] fascinating. So he's got this podcast called the Classified College Podcast. I actually think that episode is actually what I recorded it back in 2020, and it's finally getting released. And then Fabrice Mazdoor as well. Brant Dikes Joe Rhys, Britney Doe, Andrew Jones, Natalie Nixon, Christina Stop-Loss, Shannon Anderson and Eliana Lou are all lined up until the end of April, so a lot of great content coming. All right. Like I mentioned, at least an

Harpreet: [01:10:27] Episode with Alyssa

Harpreet: [01:10:29] Simpson Raw Sugar, which I think Mark, you will be very, very interested in checking out. So please do check that out. A lot of cool content and coming up. Also ask Jeremy about his what he's doing over at Netflix, but it seems like he is no longer in the building.

Speaker3: [01:10:45] He's just stepped away for a brief moment. He's going to be back.

Harpreet: [01:10:48] So just. Yes. We'll see if. See if he comes back. In the meantime, though, look like questions are welcome. Dear George, you got a question? Let me know. All right. Data George actually sent me a question on LinkedIn media. Let me address that real quick. I think this will be great because we got some great insight here. What you hear here, George, go for it.

Speaker3: [01:11:24] Typekit, thank you so much. Yeah, I said just something I sent

Speaker6: [01:11:29] Something to you, of course not. You, you take that. Thank you.

Harpreet: [01:11:35] Yeah, definitely. But this message on on LinkedIn I like you want to get some ideas for project ideas, right? So especially using SQL. So. I guess I guess a couple of good tips for you versus check out Danny Moore's content serious SQL. He's got some amazing projects that you could work

Harpreet: [01:11:58] On with SQL, [01:12:00] but

Harpreet: [01:12:00] If you are actually looking to get your hands on some real Data, Dear George, let me let me. Let me ask you if you are curious about just doing some analytics from my podcast because I got a ton of data I can get you access to. So if that's something that you'd be interested in, let me know. Because I could. I can. I can get you access to that. Very, very easily since I own the Data.

Speaker3: [01:12:29] Yes. Oh yes, I'll go first.

Harpreet: [01:12:32] Thank you. Yeah, definitely. Maybe you can help me find out what the best day of the week to release my podcast is. How about that? How's that for a project? Mark?

Speaker3: [01:12:42] Oh, I'm, of course, going to plug the Web three approach for things. There is this website called Flipside Crypto, and the reason why I like it is because even we don't care about Web three or crypto or anything like that, they make it really easy to sign up. And then they just give you a sequel kind of window pane that you can just run SQL instantly and you don't have to worry about up the infrastructure. You have to worry about Data costs. It's just a whole playground with crypto data that you can use SQL and really play around

Harpreet: [01:13:17] With, but also build

Speaker3: [01:13:18] Really cool dashboards with. So that connects to dashboards. It's on top of Snowflake, which is a data warehouse tool. And so I would highly recommend because it's real world data, it's still a little messy. We require some domain knowledge looking things up, but then you could connect it to a to a dashboard. So it's not just doing SQL, you build a data product, they can show maybe a portfolio or things. And then if you're really into it, they also have little bounties where you can use their service and the wins. You win some crypto so you can get paid to learn.

Harpreet: [01:13:52] I kind of put as well, you know, the like the bounty amounts, what what they [01:14:00] are. Mark for these equal challenges with the bounties are exactly what the bounty bounty amounts are for the skill challenges.

Speaker3: [01:14:08] I don't, but typically they've seen from like fifty to two hundred dollars worth of crypto. But it's like it's like in like weird coins. So like, I don't know how, how useful it is.

Harpreet: [01:14:19] Like, I just think

Speaker3: [01:14:21] It's cool for experience. But you know, if you want to add money to your wallet and deal with taxes, go for it.

Harpreet: [01:14:32] So, Jeremy Netflix Nasci. What's happening, man?

Speaker3: [01:14:38] It's been a crazy, crazy and concept to to be able to be there. So two days ago, I actually met the notebook infrastructure team over there. So they've been playing around with notebooks to basically get metrics from all the upstream information, so everything coming from the device to the decision makers and they are using notebooks for that. So each notebook is one metric, and we really matched on the vision because first of all, when we started Nass, we were like trying to use notebooks in production and run them on schedule to be able to get metrics and be in the comfort of the notebook and not having to deal with so much engineering. And they were having the same approach to to the to the problem. So what we learned, what I learned from this interview with them was that now that they have all these notebooks running around and talking to each other because they can do micro decision making from the notebooks and trigger action from different notebooks each together, they have a problem of templates.

Harpreet: [01:15:45] They need more

Speaker3: [01:15:46] Templates in the organization. They want people to use those notebooks on the on on any kind of scenario. So this really matches with the vision we have for Nass as being an aggregator of data science templates for people to [01:16:00] really kick start whatever they were trying to do in a minute. Like if you figure out this, this template, we want to host all of these templates into our repository and be like, Oh, I have this tool. I want to do that thing and and I can start without having to write a single line of code. I just need to understand the structure of the notebook and stop myself. So a really good, very good meeting I had with them

Harpreet: [01:16:26] And really promising

Speaker3: [01:16:28] Because now that we are here in

Harpreet: [01:16:30] Contact,

Speaker3: [01:16:32] We plan on on writing a paper about the

Harpreet: [01:16:34] Need of Data sites

Speaker3: [01:16:36] Templates for people to really jump into

Harpreet: [01:16:40] Problem

Speaker3: [01:16:41] Solving with Data

Harpreet: [01:16:42] Really quick. And and

Speaker3: [01:16:44] I hope, like a lot of things, will come out

Harpreet: [01:16:46] Because they they have needs

Speaker3: [01:16:48] Internally that could be open sourced and they have some of, you know, the community we are trying to build on. Contributing templates can also help them, but they can help also the community so it can be a good exchange of value to to be able to to work on this awesome book repository that we are trying to build. And yeah, eventually some good, good stuff will come out. Let's see. I didn't answer any question you guys may have on what I've been.

Harpreet: [01:17:16] Yeah. Like not say I like notebooks as a service. It's extremely cool. I highly encourage all of you guys to check it out. And I mean, it makes sense that Netflix is interested in it because I think they are one of the few companies that I feel like take notebooks super super seriously like they do. Yeah, they they are super all into. No, because I haven't seen any of the company that that is all in on you.

Speaker3: [01:17:39] You know, this is very interesting because what they were explaining is that and this comes to the conversation we just had because of some software engineering and where Data engineering stands and everything, they do see notebooks as as really different from the software engineering workflow. So [01:18:00] you use notebooks to create and destroy stuff very easily all the time. Like like Matrix, they are moving. So someday the wind comes this way. The Matrix would be this one, and then you want to destroy it and then create another one the other day. Because the wind has changed. You know, the the thing you want to follow on are the same, and this cannot be supported by software heavy software engineering and shouldn't be because it needs to like, be more stable, more into the infrastructure, more to the databases, more into like something that is core. So there is a loop of of using notebooks to follow all those metrics, then coming back to software engineering, building cool APIs and really strong stuff that you can rely on

Harpreet: [01:18:45] And use it

Speaker3: [01:18:46] Back into notebooks and. And this is how they play. So I learned a lot like literally it was most interesting meeting. I had my whole life really to ask.

Harpreet: [01:19:00] That's my huge hope for them. I'm looking forward to seeing what happens. Look forward to seeing it. Maybe not get acquired by Netflix. Who knows? Maybe who knows what could possibly happen, who knows what could possibly happen, but highly encourage you guys to check out Nazia. Also, check out the the project that Marc had did with Nazy. I do a LinkedIn Analytics pretty much doing what shield charges too much money for Flora for free and and it's really streamlined, super cool to use. Jeremy, how are you in San Francisco for?

Speaker3: [01:19:34] I'm there until Wednesday. I'm leaving Wednesday going back to Paris.

Harpreet: [01:19:39] All right. Well, I don't think we'll be able to connect, but hopefully, hopefully I would be back. Yeah, a little back. But it doesn't look like there's any other questions. And Nas is naras, dot, A.I. and a study I notebook as no service,

Harpreet: [01:19:58] No books, as a service.

Speaker3: [01:20:00] Harpreet, [01:20:00] do you struggle like I do with wondering why people don't prefer to come to the Boise area or the Winnipeg area over the San Francisco area? I do not understand that.

Harpreet: [01:20:13] Yeah, because it's been negative 30 degrees Celsius here for like a month and a half straight.

Speaker3: [01:20:18] 70, right? Great. I rest my case. Cold weather loving.

Harpreet: [01:20:24] I've been like, You guys don't understand how much snow. We've had like, I don't even know where to put it anymore. It's like I clean the snow. And then it comes right back, and it's just like, why what? This is so yeah, I'm excited to be back home in Sacramento. Nice 70 degree

Harpreet: [01:20:41] Weather right

Harpreet: [01:20:42] By all the breweries and wineries that are very close to my parents house, which I'll be enjoying a lot. So I'm looking forward to that.

Speaker6: [01:20:50] Let me put it this way. Tom, coming from Sydney, Australia when I moved over to London, I had 19 days straight of just cloud cover and no sunshine. I can guarantee you the first day I had sunshine. I stood in the sun for 45 minutes, going, Yes, thank God for that. The weather for some reason really has a big impact how you grow up as to where you want to be.

Harpreet: [01:21:13] Yeah, yeah. I mean,

Speaker3: [01:21:15] Margaret and I do get a lot of sun for where we're at in the in the North American continent, but yeah, it gets cold where we're at to.

Harpreet: [01:21:23] Yeah, during summertime we'll have Sun from like 4:30 a.m. to about almost 11:00 at night. It's awesome. It's quite nice. But winter sucks. Really bad. All right. Does not look like there is a lot of questions coming in. Austin, well, guys, thank you so much for joining. Thank you for hanging out. Appreciate you guys being here. Remember next two weeks? This will be Antonio taking over the hosting duties. So do show up, give him support and I'll be listening. I'll be listening. Make sure you guys are behaving [01:22:00] well. My way. All right. So let's take care. Have a good rest of the week. And remember you got one life on this planet. Why not try to do something big? Cheers, everyone.