[Bite] How to hire a great Data Scientist

April 05, 2021 DataCafé Season 1 Episode 13
[Bite] How to hire a great Data Scientist
Show Notes Transcript Chapter Markers

Welcome to the first DataCafé Bite: a bite-size episode where Jason and Jeremy drop-in for a quick chat about a relevant or newsworthy topic from the world of Data Science. In this episode, we discuss how to hire a great Data Scientist, which is a challenge faced by many companies and is not easy to get right.

From endless coding tests and weird logic puzzles, to personality quizzes and competency-based interviews; there are many examples of how companies try to assess how a candidate  handles and reacts to data problems. We share our thoughts and experiences on ways to set yourself up for success in hiring the best person for your team or company.

Have you been asked to complete a week-long data science mini-project for a company, or taken part in a data hackathon? We'd love to hear your experiences of good and bad hiring practice around Data Science.  You can email us as jason at or jeremy at with your experiences. We'll be sure to revisit this topic as it's such a rich and changing landscape.

Further Reading

Some links above may require payment or login. We are not endorsing them or receiving any payment for mentioning them. They are provided as is. Often free versions of papers are available and we would encourage you to investigate.

Recording date: 1 April 2021

Intro music by Music 4 Video Library (Patreon supporter)

Thanks for joining us in the DataCafé. You can follow us on twitter @DataCafePodcast and feel free to contact us about anything you've heard here or think would be an interesting topic in the future.

Jason  0:01  
Hi there, welcome to the DataCafe. I'm Jason.

Jeremy  0:03  
And I'm Jeremy. And today we're doing our first DataCafe bite.

I thought we weren't doing this episode for another two weeks though, Jason. 

Jason  0:25  
Yeah, we thought it'd be good to do like, like a bite-size episode where you just drop into the cafe, maybe pick up a snack, see what's going on in the world of data science right now? Or what's currently on our minds. 

Jeremy  0:36  
Okay, I like that. So that sort of presupposes that we have something to talk about. So what are we talking about today?

Jason  0:42  
Funnily enough, I thought we talked about how to hire the data scientists get data scientist into your team, whoever that team might be, and go through some of our experiences with that?

Jeremy  0:52  
Oh, plenty of experiences. No, absolutely. So I think that leads very nicely on to what makes a great data scientist in the first place. Because if you're going to hire a good data scientist, then you have to know slightly what you're looking for, which I think is a whole set of questions right there.

Jason  1:11  
Yeah. And what are we looking for? We need somebody who has, if I look at the word data scientist, I split that up straightaway and say somebody has abilities on the data side. So tech stack, coding, technical abilities, and somebody who has some sort of scientist training, can formulate experiments, run hypothesis tests, it has that tenacity that comes with going through potential failures and getting a conclusion out of that.

Jeremy  1:36  
I think you're actually right. I think there's, there's loads of attributes that people look for? Or maybe they don't realize they're looking for, and they ought to be looking for. Yeah. And then, and then there's the huge plethora of data science skills, in massive quotation marks, because if you speak to any data science manager, or hiring manager for data science they will list probably a set of skills, which are wildly divergent, and

Jason  2:04  
Yeah. Five years experience with them when they just came out last year.

Jeremy  2:11  
I think, certainly something I see. I mean, you read articles on how to hire good data scientists, and they, they all seem to suppose that you're hiring the perfect data scientist, the unicorn unicorn. Yeah, exactly. So I think we need to get a unicorn out of the way don't we.

Jason  2:29  
Yeah, and I've read articles that say it's the unicorn team that you need, which at the core of it means you need complementary skills. So you can get various aspects of scientific skills and various coding software development skills, various communication skills, and different abilities, interacting with the business or your stakeholders, and bringing them together in a team setup. And they may not all be data scientists. So defining that boundary between where as a data scientist job actually end and you're now in the realm of data engineering, or product ownership, or managerial role or leadership,

Jeremy  3:06  
I think the complementary aspect is really, really critical. You want the you want the team to behave as a cohesive team, as a unit that are going to be good at solving problems, good at producing products, and good at integrating those products into into an existing company, don't you? Because if any of those don't happen, then it's not going to be a successful outcome or experience

Jason  3:34  
Is a costly gap somewhere,

Jeremy  3:36  
Right? I think companies bring in data science talent, for many reasons, initially, but then, you know, then often we'll scratch their heads and go, why isn't it working? What what where are we missing? The the key nugget, if you like that's preventing this landing somehow. And I feel quite strongly that, that if you try and go down, for instance, the super technical route for every single data scientist you hire, every one of them is a brilliant statistician, for instance, which would be great. I mean, I'd love statisticians are terrific people to having your team, no question. But I think if you did that, at a very technical level, then that wouldn't necessarily lead Yeah, in my experience to a terrific cohesive team. 

Jason  4:20  
This gets into I think it's called the T shape for skills, which is breadth and depth. You can specialize in any one area. But it's also good to have a breadth of skills. And really, a data scientist can sometimes end up being too broadly stretched, or be too specialized in one given area. And it's nice to have that balance. And a lot of us pull this into our personal development plans to make sure I'm on top of a variety of tech stack things, but maybe I'm specifically only coding in Python, for example. And this kind of gets us to, well, it's only an individual usually in the interview. So how am I going to make sure that that individual is the person that is a great data scientist that my team needs? Or maybe just the one data scientist that I'm going to bring in? What do they need for me to get the best value out of having a data science function embedded in the company? Or

Jeremy  5:16  
 I think there are definite stages. in hiring the team. There's the sort of there's the young team. Yeah, by which I mean, not not age. I mean, small, really? And that's...

Jason  5:26  
Yeah, early in its growth.

Jeremy  5:28  
Right. Right. And that's, that's, you know, every single person you're bringing into the team should probably be contributing something different,

Jason  5:35  
Something new. Yeah,

Jeremy  5:36  
Yeah. Right. Right. And then there's the sort of more established team where you're looking for great people. And they should be great, technically, but they should have certain attributes that you're looking for, as well. Which speak to are they going to be a good fit? Are they going to be people who are going to contribute to your a team environment, and make that land? And certainly, for me, the critical thing with a data science team of any size is, that they should be exceptionally good as a team at problem solving, after discovering new solutions, discovering new models, approaches, ways of working, whatever it is, and yeah, the irony is, I think that data science holds the key for that. I mean, there is this, you know, we've covered it a few times now. And in the, in the main episodes, the concept of a reinforcement learning agent. Well, I mean, I very much see a data science team as a reinforcement learning agent, it's got it, it's got to learn from what it's done. And it's got to be able to nice trends, it's got to talk to each other about about how they've succeeded how they fail, they pass on experience, you know, all of that has to come into it. 

Jason  6:46  
It's like treating everything like an experiment, 

Jeremy  6:48  
Exactly, exactly, 

Jason  6:50  
See what's working, and where we can improve. And one of the things that this feeds into really strongly as well as an element of creativity. So it's very good to have problem solving skills that also think outside the box, as common as that phrase is, you want somebody, who can add a new dimension of thinking to a team in ways that may give a new insight to a problem that simply wouldn't have been thought of before. Because if we were all physicists, I'm a physicist, if I'm working with five physicists with a similar background,

Jeremy  7:19  
I couldn't agree more. And this is I mean, this comes straight to the heart of diversity hiring, or indeed, you know, the anti diversity hiring, lack of lack of diversity and, and how that can be so harmful in a data science team and data science hiring process, I've seen so many teams, where they've gone, you know, not not just after the same, the same gender, the same ethnicity, maybe, but also the same college, the same university, the same school, and the same degree program. 

Jason  7:50  
It's easy to relate to, yeah, you've got some understanding of the qualification or the background somehow, because of your experience with that. Yeah. And something I read, which is really interesting, is like cultural fit is one thing, but what you really want is to hire for cultural goals, find out what it is you're aspiring to be, and figure out what somebody can bring in new in a way that you haven't been able to consider. And that's really difficult. And that's why it's so great to bring in a diversity of scientific backgrounds and diversity of those technical skills and their diversity of those experiences. And if you've never heard of something before, ask about it, if such a chance to learn bring that element of growth to your team.

Jeremy  8:33  
I think the idea of a unicorn data scientist, although there are some fantastic data scientists out there can be quite harmful because data science is this massive subject, you know, aggregating computer science, statistics, you know, mathematics in a very broad sense, you know, Ai, all of this comes together. And you can't be a phenomenal technical master of all of those disciplines. You just can't Yeah, you can be very good at some of them. And you can have a passing to good knowledge of of lots of others, but but I don't think it's possible. Really.

Jason  9:07  
You can't be a unicorn. It's a myth. It's a mythical creature. Yeah, for sure.

Jeremy  9:14  
So I think I think that, you know, the team needs to support each other. And the best teams I've seen have had ended individually excellent performance in certain areas of the data science stack, if you like, but together, they present this amazing and really quite diverse sort of problem solving machine to the company into the the data driven challenges that they're presented with. 

Jason  9:37  
Okay, so what are your key takeaways Jeremy? 

Jeremy  9:39  
So I sort of take it back to the podcast that I listened to a while ago, which I really still enjoy. And I go back to now, which is a podcast produced by the Y Combinator incubator company and they view hiring people when you're a startup as this critical phase and I think for data science teams, it is a critical hire. And it's a critical process hiring new new team members. So for me, the things I look for are very, very similar to the the sort of features that they mentioned in their podcast, when they're looking for new team members, founders, even of new companies. So for me, I look for great communicators, people who can present their ideas I look for, you know, people who will support naturally instinctively support each other in a team environment. So who will lean over and go? Do you need a hand? Or I've actually done that? And I'm very happy to help 

Jason  10:40  
Absolutely. The comradery...

Jeremy  10:42  
Yeah, absolutely. It's such an efficient way of progressing a problem. And then team members who are going to memorize what they've done so that other people can learn from it. So documentation, funnily enough, it's so difficult. Yeah, you're working, you're working as a solo data scientist, you know, you can keep it in your head if you've got a big head. But if you're, if you're working in a team, you've got to be able to write it down. Yeah. And then just just the ethos of, do you seek to improve what you've done? Are you trying to iterate your ways of working, and your technical ideas and approaches? And if you bring all of those four things together, you know, it doesn't matter. Actually, what the area you're really good at, you know, I'm assuming there's also, you know, fantastic technical degree under there are some really great diverse thinking that's going on from whatever background that they've got. But actually, I think those key traits, for me come together to make a really terrific data scientist, and the rest of it will fall into place.

Jason  11:35  
Yeah. And one thing as well is you can always upskill if there's something that's potential development area. So especially in the tech stack area, and I mean, that would be my biggest takeaway is recognize when you're hiring somebody, what are the tenacious behaviours that they have a problem solving, the instinct to work well with others and really explore data be clear in their goals and assumptions, but be really tenacious and rebellious getting to the core of a problem and wanting to sell this, everything else can be upskilled in some way, you know, even right down to communication, like you said, there are upscaling seminars, workshops, things before communication needs to be improved. I've certainly gone through many of these in learning how to talk and engage with business stakeholders of various backgrounds, and ultimately as well that em cultural goals, and really, really calling out what you'd said about that similar similarity bias, that we might say, Fall Fall into, I might have 10 brilliant people, but if they're all the same, what I really need is 10 brilliantly complementary people.

Jeremy  12:48  
And I think to round that off, then whatever process you have to to pick out those key features and those those cultural attributes that you value, or the technical attributes you value, whatever process you have, it should be a process which succeeds in giving you that information. Too often, I've seen a lot of companies were putting a process in because they've seen lots of other companies doing something very similar. Sure, technical test after technical test or coding test after coding test, not that they don't have their place, just, you know, they're just pushing the buttons because everybody else is pushing them. It's sort of expected. Whereas actually, I think you'd really want to learn every step of your hiring process should be telling you something you didn't know, already should be ticking a box or crossing a box to say yes, this is a good hire, 

Jason  13:34  
Yeah tailored to what you need. So identify what it is you need as much as possible.

Jeremy  13:39  
Yeah, exactly. super. That's great, Jason, thank you very much. I've enjoyed that. 

Jason  13:43  

Jeremy  13:44  
We should do this again.

Jason  13:49  
Thanks for joining us today at the DataCafe. You can like and review this on iTunes or your preferred podcast provider. Or if you'd like to get in touch. You can email us Jason at or Jeremy at, or on Twitter at datacafepodcast. We'd love to hear your suggestions for future episodes.

Stop by for another bite!

Topic introduction
Key takeaways