image: thumbnail.jpg
Data Science Hangout | Asmae Toumi, PursueCare | Data Science Team Culture
There was a great discussion during this session starting from the question, "Is it ok to say that we are data analysts and data scientists interchangeably?"
Javier: Some of the best data scientists have the title of financial modeler or senior actuary. Some of the best programmers and package developers I know, have the title of data analyst. I think it depends on the organization. We have these guidelines that are more of an ideological function of a data scientist vs data analyst but I think in practice it’s all over the place.
Asmae: Jacqueline Nolis and Emily Robinson have good thoughts about this: https://jnolis.com/book/
(Short rules from the book I saw: An analyst creates dashboards and reports that deliver data. A decision scientist creates analyses that produce recommendations.)
Mike: I think to anyone that is OUTSIDE of Data Science / Data Analytics, the distinction is not worth discussion. If you want funding you're a Data Scientist. If you HAVE funding it doesn't matter.
Frank: No right answer (DS vs. DA)… only different ways to think about it at different times in different contexts.
Jordan: I've usually been working in a ""Science"" (science/research) domain at companies. To this end, I've sort of been the ""Data"" Scientist, maybe not ""Data Scientist""
*What about business intelligence?*
Asmae: I use data analyst and data scientist interchangeably. I think there is a meaningful difference there between a business analyst and a data scientist.
*What about data engineering?*
Asmae: We have 3 data engineers. Data scientists all want a data engineer. As a healthcare company, it’s complex to deliver the care that we do. With that complexity comes interfacing with different data vendors - different data sources so it’s really important to invest in data engineering so the reconciliation of that data across the many vendors is ensured. Data quality is also super important.
Bryan: Data engineering is the foundation of it all — need a solid data engineer to do anything
Falak: Data scientists understand the distribution of the data and figure out what class of algorithms would be suitable to handle a particular problem. Data engineers can solve the optimization problem.
Other resources shared during the hangout:
Allissa: An NIH prize challenge on maternal health that may be of interest to some - https://lnkd.in/gbtxTTcF
Bala: Top 10 Roles for your Data Science Team: https://lnkd.in/g_SVzxDcAsmae Toumi, Director of Analytics at PursueCare, to discuss the most important things going on in data science leadership.
Transcript#
This transcript was generated automatically and may contain errors.
Welcome, everybody, to the Data Science Hangout. I hope everyone's having a great week. And special hello to anyone that's joining for the first time, too. It's great to meet you. I'm Rachel. I'm the host of the Data Science Hangout. I'm calling in from Boston, as well as Asmae. I'm in the RStudio office today. This is an open space for the whole data science community to connect and chat about data science leadership, questions you're facing, and what's really just going on in the world of data science. So if you ever want to go back and rewatch past sessions or share it with someone who's missed, it is recorded.
And it will be shared to our brand new Data Science Hangout site. I don't know if anybody saw that. It's right on the RStudio page now. And I'll share it in the chat window in just a bit if our team isn't on yet. But we really want this to be a space where everybody can participate and we can hear from everyone. So there's three ways to ask questions. You can jump in live and raise your hand on Zoom, might be the easiest way with a bigger group. You can put questions in the Zoom chat. And if you're like in a coffee shop or your dogs are barking or you want me to read it, just put a little star at the end of it. But we also have a Slido link that we'll share in the chat right now where you can ask questions anonymously too. And I just want to reiterate that we love to hear from everyone, no matter your level of experience or area of work too. But with that, I'm so happy to be joined by my co-host for today, Asmae Toumi, the Director of Analytics at PursueCare. And Asmae, I'd love to just kind of turn it over to you to introduce yourself and maybe some of the work you do at PursueCare too.
Sure. Thanks everyone for joining this Data Science Hangout. I'm really excited to talk about the work that we do at PursueCare and how I fit in there, but also to learn from all of you. So hope we have some fun here. I'm Asmae. I'm the Director of Analytics and Research at PursueCare. PursueCare is a telehealth company which specializes in treating mental health disorders, specifically substance use disorder. We are a telehealth company first, but we do have also in-person clinics. We've been in operation for three years now.
Awesome. And I know you just started in this role, was it a few months ago? Yeah. I've been working at PursueCare since last year, February last year. So I'm about to have my one-year anniversary coming up. Wow. That goes by fast. So while we're waiting for a few questions to come in, I'd love to hear a little bit about how you're building up this team.
Building the data team
Yeah. So I was the company's first data person. So that came with some challenges, but also some excitement because we got to kind of explore the best tech stack for us. And we weren't, I guess, beholden to some legacy code or legacy platforms. So with that, we looked at different vendors. And given that I was an R programmer, the decision there was quite easy in terms of picking sort of the publishing platform that we were going to use, RStudio Connect, to no one's surprise there. And of course, a data warehouse that plays nicely with that and so forth.
Awesome. So how many people are part of the data team now? Yeah. So it's me and three data engineers. Yeah. We're hopefully going to grow that a lot going into next year for sure. That's something that comes up on this call a lot, is the topic of hiring data engineers versus data scientists and which comes first. And that's interesting to hear. So you have three data engineers to start.
Yeah. I think I put out this tweet, I think, a while ago. But all data scientists want is a data engineer, honestly, because we're a health care company. And I think the people that work in health care would sympathize with this. It's complex to deliver the care that we do. And with that complexity comes with interfacing different vendors, so different data sources. So it's really important to invest, I think, in data engineering such that the reconciliation of that data across the many vendors is ensured. Data quality is also super, super important in any field, but especially so in health care. So I think it made a lot of sense for us to invest in data engineers early on.
But all data scientists want is a data engineer, honestly, because we're a health care company.
Networking and community
So have the RStudio team help me monitor the chat for questions. But if no one asked anything yet, I'd love to ask you because I feel like you do an awesome job at networking and I always see you on Twitter as well and would love to hear a little bit about how you network in this data science world and if that helps with how that's helped you with maybe getting into your role now or any advice you have for all of us.
Yeah, I think my boss is on this Zoom, Nick Mercadante. He's the CEO of PursueCare. And the way we've met is through hockey Twitter. So we both like hockey. We made a hobby out of it, out of analyzing hockey data. And so the networking, I guess, there in terms of getting my job was pretty organic. And I think most of my sort of networking efforts are natural in a way. I invest in my public presence for sure in terms of time spent on Twitter. And I think over the years, if you're intentional about interacting in a positive way with the people in your community, whether that's your professional community or sports or any interests, then I think it makes it that much easier to build that connection, build that rapport. And when you're looking for your next step, then the awkwardness and all of that kind of dissipates.
Definitely. And I hope people are able to do that here in this environment too. I see Zach, you asked a question in the chat. Would you want to jump in? Yeah, so continuing on the hockey thing, I found Esme, is that how you pronounce it? Sorry if I've got it wrong. It's okay, close enough. Thank you. I found you via your Big Datadog submission last year, one that I actually really enjoyed. I believe it was the finalist as well. I was just wondering what other competitions you recommended because of course I'm new to this and I'm trying to look at other competitions. I've entered this year to the Big Data Bowl. I just want to know what other competitions were around to try and boost my CV, so like stand out from the crowd.
Sure, yeah. The NFL Big Data Bowl is such a tremendous opportunity to get exposed to a very big and complex data set. And there's a lot of eyes on that competition. NFL staff, other sports leagues have hired from people's submissions and so on. And I think that sports is a great field to kind of, you know, practice your data science skills and post it because sports fans are very much hungry for that kind of thing. I think Kaggle always has competitions ongoing and it's great because it spans many fields that you may be exploring, finance, healthcare even. So I highly recommend using Kaggle given that submissions are online. It makes it very easy to just import that code and tinker with it yourself and learn a few things along the way while also building your portfolio. Thank you.
The rise of data science video content
Craig, I see you mentioned Slice in the chat too. Do you want to ask that question live? Yes, sorry I don't have a camera. I don't know if you guys can hear me or not. Can hear you perfectly. Thanks. I was just wondering if you've noticed the data science community grow in streaming platforms like Twitch and YouTube. Obviously DRub streams on YouTube every Tuesday, I believe. I don't know if he's still doing it, honestly. But yeah, have you heard of the Slice Data Science Competition and do you plan on participating in the future?
Yeah, it's been super exciting following Slice over the summer. I think that was literally one of the highlights of my summer, quite frankly. I love just how welcoming that environment was to the participants themselves even though it was a quote-unquote competition. I learned so much from observing those participants do these timed challenges and it's been really great this kind of shift to video. Honestly, it's a little less, I think for me personally, a little less tiring to watch data science and to learn how these experts go about it live because you see them sometimes debug live or change things on the fly. It also helped a lot with, I think, my own imposter syndrome. Just seeing that they also get tripped up and they also use the pipe instead of plus whenever they're using ggplot2, that kind of thing. So yeah, definitely have noticed a bigger shift into video and it's very much a welcome one, at least for me.
Advice for the first data person at a company
Definitely, I see a few people commenting about that in the chat too. I know Julia Silgi from our studio does that quite a bit. I see there's an anonymous question that came through. A great question that's, what are some pitfalls that you would give advice on to avoid for the first data analyst in a company with no data pipeline that wants to make data-driven decisions?
Yeah, I think trying to do too much is a bad play. I think also making sure that you have a lot of face time with the stakeholders is important. I think, you know, I had sort of many ideas of what I wanted to do with this data, but when you actually sit down with the people that will be using this data, it looks differently. So I think just making sure you have that face time, that you do your due diligence in terms of, you know, what your tech stack is going to look like, make sure you right from the get-go have reproducibility and code quality in mind as you're building out those pipelines are all important considerations, I think.
Centralized vs. decentralized data teams
Definitely, thank you. Jeffrey, I see you had a question that I'd love to pass over to you to ask live, and I figured I'd say this too. If people want to introduce yourselves, I know it's been a long time since we did that, and there's quite a few people on now. So when you ask the question, feel free to introduce yourselves too. Yeah, sure. Are you all able to hear me? Yep, sounds good. I'm Jeffrey Sumner. I'm a data engineer at a gas pipeline company. My question was around, as you build out your team, do you plan to be focused more as a centralized data science team or more of a decentralized team within your organization? And if anyone else had any kind of comments on that as well, I was asked that question yesterday, and I really thought it depends more on maybe the culture of the company on how you handle that problem, but just curious on what others think.
Thank you, Jeffrey. If I may ask, what is the difference between a centralized team and a decentralized team? My definition of it, at least, is centralized would be you have more of a data science team reporting up to a data science manager, whereas decentralized would be more along the lines of you've got a team within sales, a team within marketing, or a data science member on the marketing team, the sales team, finance, whatever it may be.
Yeah, that's such an interesting question. I'm not sure if I'm best equipped to answer it. I think I'll just draw from from PursueCare. We have, of course, typical departments, marketing, finance, ops, and I think it would make sense one day to have decentralized teams such that those data scientists can become subject matter experts as much as they could in those specific fields, such as finance or marketing. I think right now it makes more sense to have a centralized team given our size. We're about 70 employees strong, and we're really at the phase of just
understanding this data and trying to better understand our business, our patients, but I think as we become a more mature company, our questions will probably become more complex in nature, and so having those decentralized teams in which they can once again have more face time with the subject matter experts would make sense, I think, but interested in other people's perspectives here.
Prabha, I see you had your hand raised. Do you have feedback to add on this topic or another question? No, on this topic, I fully agree with what Asmae is saying here. I think it is the, it can be a little bit of organizational structure. There is no right answer here. Both are valid, but for a company that is starting out a data science team, I think a centralized setup is good because they can then develop and say grow this data science function as to how it will work in this company, set some good practices, et cetera, and when you're building a team, new hires have a place to come and learn, say, company processes, how does data science work, who are the best, and so forth. However, as this function grows, having data science teams within the marketing, within HR, within finance, et cetera, et cetera, shows a level of maturity now that these functions can exist and thrive within those functions, you know, within these non-IT, non-data science organization, and continue to add value.
So, I think there is no right answer, but I think if you look at the growth of the data science function, I fully agree with what Asma says, start centrally, but as you grow, data scientists need to understand the language of the finance team, language of the operations team, et cetera, and for them to be part of that organization accelerates that growth.
Thank you, Prabha. I see a new question. Thank you, Prabha. I see a new buddy that you have your hand raised as well if you want to jump in on the topic. Yeah, thanks, Rachel. I was actually going to say everything that Prabhakar said, that when you're starting out, and this is something that we are currently struggling or working on within our company, that we are a fairly new, small, four, five people, data science team within one project. So, we are trying to hire more people, keep it more centralized, but at the same time, we've recently put out a posting for a job in a different team, because we also want to expand our work beyond our team, because our teams are segregated based on the line of work rather than just departments. So, we work on public health, but we want to do a lot more data science work in economic growth or climate change. So, we've convinced at least our economic growth team to hire one data scientist.
And the way, I think, early on, having it centralized is also very important, because there is no mentorship or guidance available from those other teams. So, that's extremely important, like Prabhakar said, to develop good practices and, you know, just organization by practices. But we are also trying to, even with decentralized, if we convince more teams to hire data scientists, to keep it more, this bits of it centralized by organizing a technical working group around it, where we need and try to standardize practices or really standard training modules.
Great point. Thank you. I think there is a happy middle, right? Once you, once you, you know, an organization has achieved this, a decentralized model. In my company, you know, the finance organization is set up in a centralized-decentralized, it exists in both ways. And I work in the supply chain department of my company. We have a finance lead for the supply chain organization. But that person, if you look at his reporting structure, he reports into the finance organization. So, he brings that finance expertise to the supply chain organization, but his manager, but his manager, he rolls up into the CFO organization, right? So, he continues to have, like, input from both places. He knows about supply chain, but he also follows all the best practices of finance within my company.
So, there is a, you know, if a data science team has grown in maturity to have enough people to live and do data science for finance and HR, etc., you don't have to cut off the connection to the centralized data science team. That can still exist.
Data analyst vs. data scientist
Jill, I see you just asked a great question right now in the chat as well. Is it okay if I pass the mic over to you, Asma? Yeah, of course. I can start reading it, Jill, but feel free to jump in, too, if you want to add any other context. But it's, is it okay to say that we are data analysts and data scientists interchangeably? Are there specific differences?
I think Jacqueline and Emily Robinson's book on data science careers has a really good definition for both and the subtle differences between the both of them. I think, and this is going to be wrong because I don't recall those definitions, but I think maybe data scientists involves just a little bit more, you know, engineering work, ML ops, DevOps sort of work. But honestly, I use them interchangeably, but others might have differing thoughts on that.
Yeah, I feel like a lot of us do use them interchangeably. I know in past conversations, we've talked a bit about data analysts more so reporting the data. I don't know if anyone has other thoughts you want to add there.
Feel free to raise your hands, too, if you want to, like, I know it's hard with 100 people to just jump in. So, Balak, did you want to add anything? Yes, I recently attended some lectures about AI ML. And in those lectures, it was explained in a very good way, the difference between data scientists and data engineers. Basically, data scientists are those people who understand the distribution about the data, and they figure out based on the distribution, what class of AI ML algorithms would be suitable to handle that particular problem. So, this is mostly done by data scientists, because there are multiple classes of AI ML algorithms that you can use to solve your problem. And that thing and that work is done by data scientists to help you by understanding your data, by looking into the data, about the distribution of the data, you can figure out what class of algorithm is best suited to solve your specific problem.
And once that is decided, then the data engineers or ML engineers, they then solve the optimization problem. So, within that specific class of algorithms, they figure out which specific algorithm with specific configurations, this algorithm, which algorithm is going to perform the best for that specific set of data.
Thank you. I see there's so many great comments coming in on the chat, too, and different thoughts that people have on this. Javier, I see your hands raised. Do you want to jump in? Hey, Rachel, I think the gentleman that just spoke kind of, you know, articulated what I was intending to say as well. But I would say that's more from like a theoretical data scientist standpoint. I mean, some of the best data scientists or, you know, our programmers or Python programmers that I know have the title of, you know, financial modeler or, you know, senior actuary. Some of the best programmers and like package developers I know have the title data analyst. So, I think it really just depends on the organization. Like, we have these guidelines that, you know, are more like an ideological function of a data scientist versus a data analyst. But I really think in practice, it's all over the place.
Some of the best data scientists or, you know, our programmers or Python programmers that I know have the title of, you know, financial modeler or, you know, senior actuary. Some of the best programmers and like package developers I know have the title data analyst. So, I think it really just depends on the organization.
Definitely. And, yeah, definitely varies by the company. I see people are saying varies based on the location as well. I think Morris brought a good question to the table. Data science versus business intelligence. Yeah. So, for me personally, I use data analyst and data scientist interchangeably. But I think there is probably a meaningful difference there between a business analyst and a data scientist.
Live coding interviews
Thank you all for this discussion, too. I see there's a lot of. Not to crack up at some of you. I see a lot of other anonymous questions coming in through on Slido. And one was going back to talking about live debugging. Someone said, I had an interview recently and they wanted me to solve write code live. I've always wanted to do that. And I was like, write code live. I've always had issues when people watch me working. It makes me uncomfortable, even if I know how to solve the problem. Do you have any tips on how to get better at that?
First off, I'm sorry you had to do that. I really don't like that sort of live interview structure. Even less so if you need to debug something live. I know that sometimes I'm on Zoom meetings where I'm displaying a dataset or a graph and sometimes have to code things on the fly because someone asked a question. I make tons of mistakes, tons of typos in front of everyone. That's just the nature of it. In terms of getting better at that, if that's going to be sort of the standard for some companies to still have those live coding interviews, maybe streaming might be a good avenue to do that. You could put up a stream and have your buddies join, sort of a low-pressure way to do that. Would I do it to get better at live coding interviews? No, because I just wouldn't attend those. But streaming is tons of fun and I think Tan can speak to that. I really enjoy his streams and learn a lot in a low-stakes, low-pressure way.
Tan, do you have any tips to share with us then? Do I put him in the spot? No worries. You can put it in the chat too. I'm sure everyone has their stories, but I've interviewed people who have bombed technical interview back when I thought that they were a good idea. I've hired them anyway and they are awesome. Great employees, great team members, great across the board.
Great. And just while we see Frank on video, I feel like we also say congratulations on your awesome marathon time as well. I just saw that on LinkedIn. But Sep, I see you had your hand raised too. Yeah, I was just going to kind of add something that I've learned that was a bit interesting, is that sometimes the way to kind of... I don't like live coding interviews whatsoever, but if you were to participate in one, sometimes it's not about the right answer or being able to find the bug or whatever it is. Sometimes it's just a matter of seeing how someone reasons through doing it or how they approach the problem. It's similar to kind of like those really random consulting interview questions that you might get, like how many jelly beans can you fit into an airplane or something. There's no real right answer, but it's just a matter of how you reason it.
So there could be some utility there. So I think that there may be value if you can kind of change your perspective on what it is they're trying to do and maybe helps calm you in that particular moment, that there's not necessarily a right answer. Sometimes it's just a matter of how you work in that kind of uncomfortable high pressure or whatever situation. So not endorsing them, but something that you might be able to kind of mental jujitsu to help you make yourself a little bit more comfortable.
Thank you, Seb. That's a great point. I hear somebody was just about to jump in. Prabha here. We do live coding interviewing for recruiting into our team. And I agree with what Seb says, right? We are, even if they're not able to do, execute or write code that runs flawlessly, that, you know, there are, I would, in those coding, in the interview session, we evaluate them on seven or eight different categories. Writing code is one of those categories. So just like how Seb said it, it is their way of thinking. Do they, and one of the
evaluation criteria is when we give them a task, let's say, do some exploratory data analysis with the dataset. Do they ask, do they engage us? We are encouraging them to engage the interviewing team and ask the kind of questions that they would try to get to the answer that we are asking them. So that level of engagement, are they asking the right questions? Coding is one out of, say, seven or eight different criteria we're evaluating in that interview session.
But now I, you know, based on all the feedback and the sort of resistance I'm seeing to live coding, I, I'll ask some of our team members who've gone through this live coding recruitment, who are part of the team, I'm going to go back and ask them, how did they feel doing the live coding session? I, you know, it's an interesting topic. And thanks for those comments.
Yeah. That's a good idea to go back, go back to them and ask them about it. Asma, I, oh, sorry, go ahead. Oh, I just remember an interview that I, that I very much enjoyed in which I was given a take-home challenge and then was brought back for a live interview with the team to discuss my approach in depth, in which they also had the opportunity to ask questions about my approach and also give constructive feedback about that approach. And I think also it's a good way to sort of gauge, you know, emotional intelligence. How do you take that feedback and test for the, for the soft skills? So I think that's, I think that's a, that's a much happier medium that I think makes as many people comfortable given that, you know, people do struggle with, with live timed challenges.
Imposter syndrome
Definitely. And I think this question kind of ties in to this discussion as well. But I see Hugh, you asked any advice on dealing with imposter syndrome? Yeah. I'm Hugh Welch. I'm a management analyst with VHA. You mentioned imposter syndrome earlier, and that's something I'm always interested in hearing about, how people have dealt with that and any strategies that they've developed.
Yeah. I think imposter syndrome is kind of just a part of the, part of the deal here in our fields. But I think, you know, surrounding yourself with humble, capable friends, whether that's through Twitter or Slack communities, in which there is that honest, sort of, rapport in which, you know, you share your wins and successes, but you also share your struggles. And there's an environment in which you feel comfortable asking even the smallest of questions, or so you perceive, I think goes a long way into dampening that imposter syndrome. And communities such as this one, like we're here, you know, sharing our struggles and so on. So I think, yeah, I think investing in sort of a group of friends in a professional capacity goes a long way in terms of helping that.
And I love talking about it in this type of format as well, because you realize, like, how many people do have imposter syndrome. And what has also been a little eye-opening to me is a bunch of data science leaders as well that are on here who will, like, message me after the fact and say something about imposter syndrome, too. So I think it's good to recognize that a lot of us are feeling that.
That's the thing that always surprises me the most. When you talk to somebody who doesn't have a name for what they're feeling, you know, like the idea that somebody's going to come along and say, you don't belong here. Like, I feel that. I thought it was only me. Like, nope, that's everybody. Good luck.
That's the thing that always surprises me the most. When you talk to somebody who doesn't have a name for what they're feeling, you know, like the idea that somebody's going to come along and say, you don't belong here. Like, I feel that. I thought it was only me. Like, nope, that's everybody. Good luck.
Live coding vs. take-homes
Tan, I see you have your hand raised. Yeah, I was asked for some thoughts on streaming and live coding at one point a little bit earlier here. Personally, I think that live coding interviews get a bad rep in that I want to see how people think. And I try to stream the same way where I don't present myself as an expert, even though I consider myself a good R coder. I work through projects on stream, and I share how I think, and I share messing up. And I think being comfortable with that is useful. I think my problem with take-homes, I'll counter the massive support for take-homes, is that I feel like often it ends up being that people spend so much time doing them. Like, for me personally, I would rather sit, and this is coming from a perfectly biased background, because I am a streamer, and I'm very comfortable sharing. But at the same time, I would rather respect someone's time and give them the option to do a live coding interview, because really what I'm interested in isn't, can you deliver a perfect model at the end of the interview, or solve this cleaning, or whatever. I want to see the process, and I want to see what it's like to work with you as you talk through stuff and what you think. And so when you do take-homes, if I give you a week deadline, and I say spend 24 hours on it, you're still going to spend all week, like nights on this thing, trying to get this job. I don't love that.
And if you can build up fluency and confidence, even in just pair programming environments, I think that means so much that, you know, it matters that you can share your process, and that even the best of us, Eric is a great R coder, and, you know, we'll all struggle on stream, and feel safe doing so. And there's a huge part about psychological safety there, and building up your own safety, and build up your own comfort. And even if it's just pair programming, finding a friend, finding a community that will help you do that, I think that matters a ton. Streaming, definitely my preferred way to do that, but there's definitely more than one way to skin that cat. Thank you so much, Tan. That's really helpful.
I see there's quite a few hands raised on this topic, too. I'm trying to go in order. I see, Mike, you have your hand raised, if you want to jump in. Yeah, thanks. I mean, I suffer from impostor syndrome. I think it's inevitable, and it goes with the territory, I think, because we become specialists in a very narrow field, right? And the more specialist you become, the narrower the field it gets. And then eventually, you hear someone who you think is in a similar field talk about things that you think you don't understand, and that pulls the rug out from under you, and, you know, you're bound to feel that, oh, my goodness, I don't know anything, right? So I think it's especially tough if you're trying to lead something, right? Because then you have people working for you in your team who are the ones who are pulling the rug from under you, right? So I think you've got to try and, you know, develop a little bit of a thick skin, but also know what you know, know what you don't know, and just be able to say, look, I'm sorry, I don't understand what you're telling me. You know, help me out here. Help me learn.
Definitely. Thank you for sharing that, Mike. I know there's a ton of other questions, too, that I want to make sure that we get to as well, but I'd love to hear the thoughts on this. And Anubhati, I see you had your hand raised, too. You want to jump in? Yeah, I'm also going to add on to this hot topic of life coding. My personal preference is always take home. And one of the things that when we send out take home exercises, or when I do them myself, we try to focus on asking candidates to comment their code really well. So make sure that we are able to see their thinking process. And that's something I also try to practice, that I comment on the code, why I'm doing X, Y, Z, and, you know, what is my thought process behind that code. So I think that that part can be achieved through commenting. You can definitely pull a piece of code from Stack Overflow, but your comments will show why you're doing something. So, yeah.
Learning SQL as an R user
Definitely. Let's see, I want to go over to one anonymous question as well, and make sure I address some of those, too. But definitely leave your hands up as well, so I can come back over that way. But Asma, one of the anonymous questions was, I use R to query databases, but I've noticed SQL is common in many industries. Wondering what are your views on learning SQL when I can do the same in R? Very spoiled as R users with the ecosystem of packages that can make it quite validating for you to sidestep learning SQL. However, I found it to be a very useful skill that I personally should invest time in, and I think to generalize that other people should, too. Just drawing on my own experience, so I'm the sole data scientist with three data engineers, and they write everything in SQL. And so, you know, to better understand, you know, their code and how that translates to the data that they're feeding me, I think definitely having an understanding, just being able to read it is helpful without, you know, necessarily having to wait for them to explain it to me. So I think definitely investing in SQL is only a win for you personally, and it's good for your resume as well to have that.
investing in SQL is only a win for you personally, and it's good for your resume as well to have that.
That's a helpful point on working with others, too. Craig is exposing me right now. I did see that tweet. Federica, I see you had your hand raised as well. I'd love to have you ask your question. Hi. More than a question, I meant to say something about this impostor syndrome. As I said, my thoughts are like thinking about being a student and going to university and following different classes. So you might be able to be an expert and pass your exam because you stop everything and you focus on the exam and then you'll be able to release all your learning for that task, passing the assessment. But while you're learning and you're following different classes, you won't be able to be an expert on all the topics. So while you're doing life coding and you post your resume for a job, you might be aware of different things. You have a different sort of skills, soft skills, hard skills. But when someone calls you for a job, they may ask you something that you are not doing very frequently, but you are a knowledge of. And so the time that it takes for you to solve that task is maybe a little bit longer than what it would usually take if you are doing that task every day, as I've seen.
So when you work, you do a routine. You are positioned inside a company doing something and that something is a work, so it's repetitive. You won't be able to span on everything. You have colleagues, you have experts to rely with and ask if you need help. So the difficulty is when you do an interview and with the life coding, you must be sort of prepared for that task. Because when you work, you do it routinely. So that this is my, that's what I was thinking. And I found quite difficult sometimes to be able to do everything very fast. So thank you.
Thank you, Federica. Eric, I see you had your hand raised too. Well, first of all, Tan is way too kind. I'm not nearly as proficient at coding and live streams as he is, and he's rescued me more than half the time on debugging and all that. But I think the part I struggle with, especially as I start to get more involved in my organization, being asked to maybe evaluate candidates coming in and things like that is finding, like I'm not allowed to ask for live coding, like that's off the table in our process. But I've always struggled with finding ways to assess how does somebody go about solving a problem and their techniques, maybe not necessarily getting the right answer, but what investigational traits do they have to solve things? Everyone knows, that's why I watch me before, I don't
know the answer to everything. I just have a few selected sources I look at and I try things out and I make notes. But trying to assess that in these evaluations is just really painful when you only have like 30 minutes to 45 minutes maybe with a candidate. And it's not so much a question, just kind of just goes ranting about, I wish there was a better way to understand how people go about solving problems and not be misled by just looking at the end result of like a GitHub repo or a take-home exam or whatnot. But it's been a healthy discussion, I've enjoyed listening.
Yeah, definitely. Asma, do you have any other thoughts on that?
Um, I remember complaining to my organic chemistry professor about his time test because there were so many questions it was not feasible to get to all of them. And I just remember him asking like, what is the point of that really? Like, if I was, let's say, you know, trying to, trying to, I guess, like land a, land a job as like a ICU, like trauma person, then it makes sense to have these time tests in which you're like very rigorously evaluated on the end product. And if you've, you know, caught on certain things in the assessment. So I think I, yeah, I was like thinking about that as people were talking. And I think we just, I think for people that do give these live coding interviews, I think just like really, really being thoughtful about what you're trying to test and the end goal here could make for more holistic interview process that really gets at what you're trying to understand about your candidate. Yeah. Sorry, that was a random example, but no, that was, that's extremely helpful.
Convincing clinical stakeholders to trust data
Yeah. I, I saw there was a early, a question earlier for you about, so it says in a healthcare company, how do you convince stakeholders who are clinical experts, for example, doctors to trust you and your results as the data expert? Yeah, I think I've been super fortunate in that regard. There wasn't much convincing that I needed to do to convince the stakeholders of the value of data. Maybe that's because we are a digital first company. So, you know, yes, we do have physical clinics, but, but everything is, is data. And so, you know, because everything is, is data and if there is, you know, an establishment of, of, of trust there in terms of you producing quality work and, and reliable results, then that, that part is pretty seamless, I think, in trying to convince the stakeholders of the value.
There just isn't, there is no question as to like the, the value of the data, because it's something that the stakeholders need to look at pretty much every day for, you know, assessing patients, assessing processes and identifying how those can, can be improved and optimized. Maybe that has to do with the culture. So Nick was on earlier in the call, I think he dropped off, but, you know, at every monthly team-wide meeting and really every internal meeting, you know, he, he asks for departments to give an update and he, he puts into words a data-driven update. So it's something from the get-go that everyone is expected to look at and to ask for and to interface with me and my team to obtain those insights.
So very grateful to have that, that culture and, and for my job to be that much easier in which I don't have to really sell people on, on using, on using this stuff. Really the, the convincing is just in terms of, you know, how the data should be best displayed for them to then be able to use it to, to make decisions. So we'll have conversations about, about that, but really it's, it's been, it's been almost too easy for us in that regard.
Assessing data culture before joining a company
That's a great point about culture too. And I, I'm thinking about people who are looking for new jobs now. I know we're talking about like interviewing techniques and maybe that's one way to check for culture, but what are some other ways that you can get a better feel for what that culture is going to be like before you start there? Yeah, I think, you know, maybe asking them about how they think they would utilize the, the data science team, how they're currently utilizing the, the data science team and, you know, making sure that, that, that data science team isn't just, you know, off to the corner doing their own thing, but it's fully integrated within, within the company.
So we talked about like decentralized data science teams versus centralized data science teams. So if it's decentralizing, you have, you know, a data scientist into every department. Well, then that's, that's kind of your answer, right? They've invested in that. But I think, you know, in, in certain fields, you just drawing from, from what I know about sports and my colleagues in, in sports, you know, many teams have invested in, in data scientists because, you know, teams think that that gives them a competitive advantage, but in terms of actually utilizing the findings of that data science teams, you know, that that's been lacking, there's many reasons for that. And it's, it's, it's certainly not the same reasons across every team, but, but it's, it's, it's, it's fascinating to look into. And I think some people have thrown around, like, you know, maybe we need decision scientists on top of having a data scientist or data science teams. And I think that's, that's a really interesting conversation to have.
Decision scientists and data integration
Is that something like a decision scientist? Is that someone that you would probably hire into your organization? I think it would make sense. Yeah. Moving forward. I think right now at my company, we're really at the early stages of just building that, that infrastructure and, and, and doing QA on that. And at the same time, of course, delivering insights and value. But yeah, I think, I think as we mature, you know, a decision scientist may make a lot of sense, you know, because they're really the, the expert in terms of taking that data and then helping, helping to improve processes, maybe clinical processes. So I think that that would make a ton of sense actually to, to, to invest in going forward.
I'm really interested actually in hearing if you are in a company that employs these decision scientists or these sort of operations engineers that make use of data specifically for process improvement. That's, I would love to hear from you and what that looks like.
I know Frank, your team had changed the title to decision intelligence. I don't know if you have options, if you have things to weigh in on there. Yeah. So I'll just say that it also blends into our conversation about data scientists versus data analysts. The analysts that work on my team within Target, within supply chain, particularly in transportation, we are interested, like we are a centralized team. We sit with a lot of other data analysts, but my team, my analysts are so embedded with our stakeholders and our users. They know the business almost just as well as the business does, and they know how to manipulate data and model data and pull data just as well as any of the other centralized teams. So I use the term full stack data analyst. It is hard and it is rare. And maybe the learning process is very draining for some folks, but we get a ton of value on both sides of the table and having people who are, let's say 70 to 80% good at a lot of different things.
I have something to contribute to that. I mean, the term decision scientist, I've heard of it, but my challenge to that need for such a position, I challenge that notion primarily because if let's say you're presenting analysis and data science work to a finance person or to a supply chain functional person, they are the decision scientists here. They are the ones to make a decision. So if there is a data scientist and then there is a decision scientist, to the finance leadership person who needs to make the decision to define course of action for an organization, whether the position is called decision scientist or data scientist, they are still the outsider. So the decision makers already exist in organizations. They are the owners of business processes and P&Ls and so forth. I think of it as we need to cultivate the business owners to become better decision makers with the data that we provide them.
I think of it as we need to cultivate the business owners to become better decision makers with the data that we provide them.
I'll make one last comment there. So it's a great point, right? At the end of the day, the analysts are not the decision makers, but the fact of the matter is they become partners at the table because they are able to really understand the decision, any decision frameworks that are going to be put in place in the iterations as well as say, hey, that's great, but the uncertainty in the data is this. And we know that a lot of business users want to use point estimates to make their forecasts or their decisions. Having the trust built between the data people, call them scientists, analysts, whatever you want to, and then people making decisions so that the data people have a lot of weight and input on the decision, I found to be, it pays off. I see Christy put a great comment into the chat too that I think that all goes back to culture too.
Organizational culture and maturing data science
Yeah, I mean, it really does depend on not just like team culture. For example, it sounds like Frank has a really great culture that he and his team have created at Target. You know, the way that you represent your team by how you conduct interviews and things like that, that's kind of team culture. But then you have your company culture, which is really a system of systems, right? So you have a lot of different teams that are producing and consuming data. And I think that some mechanisms of that workflow are not going to work across all organizations. Another way of saying that is organizational structure plays a really big presence in how data science and data analytics get consumed and communicated across an organization or a company.
You know, a lot of that has to do with staffing, how many data scientists there are. I think, you know, I work in experimentation. I didn't actually introduce myself. Hi, I'm Christy. I work in experimentation at Grubhub. Experimentation is a really great example of where some of these communication type challenges come to a head because experimentation sits at the intersection of finance and product and marketing and engineering and data science and analytics. And so we all have to really think hard about how to work together. And it's not easy. And I'll also say, you know, I don't think a lot of even top tier organizations have this problem fully fleshed out and solved.
For a lot of the newer data scientists on the call, I think it's really important as you take on your new career as a data scientist to just kind of keep in your back pocket. Data science is still really, really new relative to software development. And developers have put into place all kinds of processes and workflows over the years to operationalize their work very efficiently. And I think that that's still kind of a source of frustration for a lot of data scientists in industry. So, you know, as you take on your new job, exercise a little bit of patience. Remember that, you know, not all of this stuff is really fleshed out yet. And you can be a source of inspiration for solving some of these problems.
I know that that's something that's really driven me as I became a data scientist just five or six years ago, is I'm super passionate about how do we operationalize data science effectively? Because I would say that most companies are probably getting about 60 percent of their ROI on the data scientists they hire. We can do so much more than we're, I think, able to do today. We just need the tools and the leadership and the, I think, just the experience even. We have to fail to learn. And so, like, how do we figure out how to move these things forward? It's a big space. But I think, you know, the maturity is coming. We just we need more people like you. Sorry, that was that. I love that. I'm going to hide again. No, I mean, I swear we were
we were just saying we need to save that little snippet too. That's super helpful.
Asma, I know I just messaged you to make sure it's okay to stay on a few more minutes because there's so many other great questions, too. And I feel like an hour just went by so quickly.
But, Craig, I see you had a great question around remote work. Do you want to ask that one?
Remote work and career advancement
Sure. Yeah. So, I work for a consulting firm and the client is all on premises. And, you know, I've been hired to work remotely. So, yeah, I just wanted to see if anybody wanted to talk about how hiring remotely has gone and whether or not you see that remaining in the field or if you if there's some chance that we're going to be going back to more in-person type situations, because I don't think a hybrid environment works from an advancement, a career advancement perspective. I see, obviously, you're at work, but I don't know if your colleagues are or not.
So, if anybody wants to talk about that more, feel free. Yeah, I think, you know, very fortunate once again that that PursueCare was created with the intention of having a significant portion of its workforce be fully remote. And so I think right off the bat, the culture and the tools we've adopted and the processes we've adopted were friendly to that, to having the workforce be remote. I think there's a lot of interesting debate about sort of career advancement for people that are remote versus in-person, that people that are remote may miss out on promotions or not be considered for those. I think that's a valid conversation to have with your managers or just be mindful of that if that's your goal within the company.
I think, you know, there's pros and cons to being remote for myself. You know, do I wish I had more face time with other stakeholders sort of in-person, you know, grabbing lunch or doing all that. But I mean, the pandemic is a thing. So that's not, I mean, if the pandemic was over, let's say, then yeah, I wish I would have more of those organic experiences for bonding and all of that. But I think being remote has paid dividends in terms of my productivity and my overall happiness and my ability to kind of structure my day. I know that personally, I don't really like to work in mornings. My productivity kind of starts at like 11 a.m. until 3 p.m. and then I have to go make lunch and eat and then I like to kind of get back into it in the evening sometimes. So that flexibility has been really great for me personally. But I think it does come with some drawbacks that I think people should be mindful of.
Does anyone else have any other thoughts here too or want to jump in? I don't want to switch the topic too quickly. You can go first. Go ahead. So I've had my last couple of jobs now remote and all my customers that I serve were scattered all over the United States. So there was no possible way that I was going to get FaceTime with them. I sat in Boston, but my customers were in St. Louis and California and Chicago and wherever else. I could have been on the moon and it wouldn't matter. My wife works for one of the largest companies here and same issue. She has people all over the globe that she works with on the team.
There's no way you're going to get a FaceTime. And if you're requiring the FaceTime for people to advance, then that means you don't value the work. The work should speak for itself. And if it doesn't, then I think as a corporate culture, you're not valuing it enough. Because most of this stuff can be done remotely. It's not physically possible to be in the office with everybody you need to work with. It's just not possible, especially with these globalized organizations.
The work should speak for itself. And if it doesn't, then I think as a corporate culture, you're not valuing it enough.
And to Azmeh's point, it gives people the flexibility. At the end of the day is I only care if you get the job done. I don't care if it takes an hour, 30 minutes, or 12 hours. We all know the schedule. We make it realistic. And that's it. It's hit the deadlines. I worked remotely for a European company. And all they told me was, you need to get these amount of documents in by X time on Friday. And that's it. And some days I would be scrambling late Thursday night because it wasn't going that well. And then one time I delivered something early and the rating agency was in their office. They were like, hey, the rating agency was asking all about the document. I said, yeah, we just got in. This is an example of it. The rating agency was great. You get an A in that department. So it was pretty clear, just make this deadline.
But that's my two cents on that.
Right. For sure. Yeah. I couldn't imagine kind of sitting at an office from nine to five. It's just, it doesn't really drive with how most people work. One thing, though, that I think about, I saw this on Twitter, someone saying, you know, junior software engineers, because of this shift to remote work, will have their learning stunted or something like that because they just don't have that much face time with more senior software engineers in which they can kind of just like pop into their office and ask questions or see how it's resolved. And that's something that I think about maybe when I will hire more junior data scientists right out of school.
I think, yeah, I think as long as there's structure in place to make sure that there's that there's enough sort of one on one time and mentorship time that we can sidestep that. But that's something I think about as well with the shift to remote work. I know that that was really valuable for me when I was a data analyst at Massachusetts Gen. I would just like constantly pop into my more senior people and ask them to help me out. And I think it's just more awkward to do it with maybe not more awkward, but I'm slacking just like, hey, like, I'm stuck. Can you get on zoom with me or something? But it's just a little, I guess a little more awkward to do it that way, versus just like barging into your co workers office. But yeah.
Yeah, I know some people have talked about like having office hours to I wonder, like having virtual office hours helps with that. Yeah, definitely. Zach, I see you had your hand raised. Yeah. So I was going to the point like, I think I may have said a lot about me being junior at the moment. So I'm on a bit of a different thing. I'm on a training. And so I went into the job with no skills. And then they teach you some skills. Even then I'm the only person in my team, I think. So if I have a problem with something, I have to Google it, or like, it's mainly Excel or Microsoft Power BI. So I'm not sure if Tan's still here or not, but like for personal projects, I talk to Tan quite a lot. And he answers like, he's that person basically that I go to if I have any questions. So yeah, I think, well, I think I would benefit from going to the office and just having, where it's less awkward. I'm not like sending a message to someone. It's also one of these things where you can have those like two minute quick conversations. You can't really have a hey, are you through on Zoom? You don't know if they're through or not. They could be going for a meeting. And it's that thing that I think, like that timing where I think it's actually critical that a lot of people don't realise. It's the end of meeting, get into next meeting, those five minutes. So you can answer a lot of questions for the junior person where you don't get that in Zoom or Teams or whatever platform you're using. Yeah. I like walking in the hallway with people all the time.
Professional development time
Mark, I see you had a great question you just asked now on professional development. Would you want to jump in and add some context to that one? Yeah, sure. And I guess it's part of the question and part of a statement that I think it's good to have this sort of scheduled time. But I asked if anyone else has scheduled time to work on non-work projects. I saw that you had a question to work on non-work projects. I saw Brian responded that he designates some portion of Fridays for folks on his team to work on developing technical skills. And I know that's something, I guess, again, a statement that I've found to be helpful is that I have that same sort of time that I've gotten my manager sign on for working on non-work projects that I find interesting to help develop skills that I otherwise wouldn't necessarily encounter in day-to-day work.
So I find it useful. I just didn't know if anyone else had any thoughts on that sort of thing to work today. And I know also there's sometimes people suggest here, oh, you should code all the time outside of work. But I know some people don't like to, they want to have a clean separation between the work day and then free time on afternoons and the weekends. And that's a good way to still get in personal practice without having to sacrifice free time.
Yeah, that's a great question. And Rachel's comment is perfect. They said, we're told, feel free to take time for professional development, but it's implied that is when your work is finished and it never is. That's perfect. I think I'll add that once again, culture is important. And I'm not just saying that as a throwaway. So Nick, my CEO, evaluates his staff based on what's called SMART goals. I can't recall for the life of me what that acronym is, but SMART goals is things that you've achieved over the course of one year. And one of those SMART goals has to be the demonstration of your professional development. And that could look differently for people. And so having, yes, exactly, specific, measurable, achievable, relevant, and time bound. And because of that, because it's literally part of my SMART goals, and that justifies time that I set aside for my professional development.
Yeah, and it doesn't have to be coding. Professional development could be literally reading, I don't know, papers on plant science. Oftentimes, inspiration and the learning of new techniques and new frameworks could come from fields that could look very different from the field that you're working in currently. So I think having sort of leadership that understands that, and that accommodates that, and that embeds it in your performance is really great.
Oftentimes, inspiration and the learning of new techniques and new frameworks could come from fields that could look very different from the field that you're working in currently.
And so with that, I told Nick, I wanted to TA a data science class Fridays from 10 to 1pm. And you know, I didn't really need to convince him of that. He's like, okay, yeah, that makes sense, because you're going to learn from your students. And you're going to learn, you know, about how to communicate, you know, things to people that don't have experience with doing that. And that's a big skill to have, because you'll be, I think, as a data scientist, often interfacing with people that don't understand it, and that you'll need to explain it. So yeah.
I love that. You're TAing that class, too. That's awesome.
I want to try and get to everyone's questions, too, but I know some people had to drop. So if you're on, and there's a question that we didn't get to yet, feel free to raise your hand and ask that live, too.
Managing a data engineering team
One of the questions from earlier, Asma, was, I know we talked about how you have three data engineers, and then data scientists. And if that's all under your team, as well, like, do you manage the data engineering side? And the question was, is your area of expertise also in data engineering? If not, how do you go about managing that role?
Right. It's definitely not part of my skill set, but I'm learning, you know, slowly but surely through them, you know, good practices and so forth. So I think I'm, like, a pseudo, really pseudo data architect in that way. So, you know, I sort of interface with the vendors in terms of interface with the vendors in terms of negotiating, you know, data access and how to obtain that data. And then I go back to my data engineers and tell them, okay, this is what the data looks like. This is, you know, the cleaning that needs to be done. This is how we should format it such that it makes it easier to, you know, slice and dice the data whenever we need to. And so that's how I've managed this team is sort of, you know, give them an understanding of the business, which then informs how the data should ideally be structured to meet those business needs.
But yeah, I certainly don't really have intimate experience of data engineering. I have some ideas,
but it's definitely them almost managing me in a sense. So it's a good two-way street there.
Definitely. I see one other question from Yuri that we didn't ask. And I know we're quite over, so we'll start to wrap up in a few minutes here. But the question was, how much of data science is data exploration versus machine learning versus deep learning?
Yeah. So I really think it depends on the company and the team and so forth. I can tell you, PursueCare machine learning right now occupies zero of our time just because we're just literally not there yet. But it's something that, you know, once our data warehouse is mature and we've understood the basics about our business, then for sure that will become a bigger part of our time.
Can I elaborate on the question?
Yes, definitely. Sorry, I didn't know you were still there.
Yeah. So when I try to learn this material and I feel before going into machine learning, which feels much more complex, I try to do the data science part without it. So I wonder how you study these layers of complications. Am I clear?
I think so. You know, I think having exposure to machine learning is important, especially if you can fit it into your day, because there will be a time in which you're probably going to have a question that involves you, you know, modeling and maybe drawing from a machine learning model. But yeah, I think it's kind of the things like learn to, what's the expression, learn to walk before you run type of thing. I think getting really good at using group buy and summarize will make your stakeholders very, very happy. So I think if I were to give advice on someone kind of starting out, I'd focus on getting really good at, you know, being able to slice and dice the data and represent it in efficient ways before, you know, moving on to more fancier and more complex questions. Thank you. Yeah.
I think getting really good at using group buy and summarize will make your stakeholders very, very happy.
Wrapping up
Thank you all so much for joining and staying on past the hour too. I really appreciate you jumping on here, Asma, and sharing all your insights and experience with us too. What's the best way for people to get in contact with you if they had follow up questions?
God, this is so fun. I could literally do this for hours, but we all have jobs to go to. Yeah, I think the best way is Twitter. I'm usually very active on there and will definitely respond. So thanks. I'm known as Jeep Wrangler owner. It's yes, it is my cow.
Asma tweets about all things, data science, jeeps, cats, snowboarding, I've noticed.
I forgot we were going to do this from a brewery as well. That was the goal. Oh, yeah. But Omicron. Yeah, one of these days we'll all be in person. Sure. Thank you so much. And thank you all for joining. Have a great rest of the day.