#77 Acing the Data Science Interview

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com.

You're listening to Data Framed, a podcast by Data Camp. In this show, you'll hear all the latest trends and insights in data science. Whether you're just getting started in your data career or you're a data leader looking to scale data driven decisions in your organization. Join us for in depth design discussions with data and analytics leaders at the forefront of the data revolution. Let's dive right in.

Hello everyone, this is Adele, Data science educator and evangelist at Datacamp. I'm excited to kick off our first episode of 2022 not only with a fresh look and feel, but with a discussion with Nick Singh and Kevin Wo on their latest book on acing the Data Science Interview. You know, as we enter the new year, it seems like we're telescoping into the future of work companies embracing remote work, the Great Resignation putting pressure on teams to create more fulfilling roles signals and expanding opportunity for applicants to find their dream roles in data science, but also for hiring managers to create awesome candidate experiences. And there are no better people than Nick and Kevin to discuss these exact topics. Nick Singh started his career as a software engineer on Facebook's growth theme and most recently worked at Safegraph on a location analytics startup. He graduated from the University of Virginia with a degree in Systems Engineering and a minor in Computer Science and Applied Math. In college, he interned at Microsoft and on the Data Infrastructure team at Google's Nest Labs. Kevin Wo is currently a data scientist at a hedge fund and previously was a data scientist at Facebook, working on Facebook groups. He holds a degree in Computer Science from the University of Pennsylvania and a degree in business from Wharton. And in college he entered at Wall street, at Facebook and Bloomberg. He's also a Data Camp instructor. Now let's dive right in. Nick, Kevin, it's great to have you on the show. I'm excited to unpack your latest book on acing the Data Science Interview, Best practices for applicants and hiring managers and all of that fun stuff but before we get started, can you tell me a bit more about your background and how you got into data science?

Sure. So glad to be on the show. Actually, Nick and I met in high school. We studied biology together and even ran a club. So it's kind of a little bit of a somewhat unorthodox path for me personally. I actually switched into computer science upon going to college and ended up learning some stats and finance as well. So kind of was trying to learn as much as possible about different fields and switched from software engineering to data science. Ended up working at Facebook doing data science and then also a hedge fund as well. And then while I was in New York, sort of bumped into data camp and was also an instructor for a class there. So yeah, just been pretty, pretty immersed in the space.

Yeah. And my background's similar. So as I mentioned, Kevin and I, we grew up together. We were actually roommates when we both worked at Facebook. Him as a data scientist, me as a software engineer on the growth team doing a bunch of AB tests and data driven experimentation. So not quite a data science role, but like something that was very data driven as a work. And I've also held different roles in data infra at Google and worked at a data alternative data company most recently. But mostly I'm a career coach and that's where my passion lies, helping people with careers. So Kevin and I really just bonded over the fact that like, why is there no cracking the coding interview for data science? You know, where's the leetcode for data science? Why is this field so confusing? So that's what we worked on for the last year. We combined Kevin's experience being a data scientist at Facebook and Wall street with my own experience being a software engineer turned career coach with stints in different types of data roles as well, and data companies to really come together and make this book.

That's great. And this book was definitely a decade in the making. I found the book to be a great resource covering the entire gamut of the interview process and preparation for a data science role. I don't want to spoil the book, but the book is broadly divided into two large sections. One on really the behavioral side of acing the data science interview, as in how should aspiring data scientists approach the job hunt? And the other one on the technical side of acing a data science interview. I want to unpack the first section now, but before I'd love if we can settle this once and for all, what should go first on a resume, work experience or education?

Yeah, that's A great question. And like all these sort of questions, man, answer is it depends. I know, I know. That's so unsatisfying. But here's something to really think about that will answer a lot of your resume questions. Right? We've got 10 seconds to read your resume. That's how long a recruiter or hiring manager is looking at your thing for. And we read top to bottom, left to right, in English. What does that mean? You got to get your most important stuff up there on the top, visible, obvious, so that someone just looking at it real quick can understand. This is Adele. Here's where he works. Or the most impressive thing about you. Right, because we're trying to impress them in 10 seconds. So what goes first, work experience or education? Well, whichever one's more impressive. If you went to school at mit, but you haven't really had work experience, let's go list MIT first. Or maybe you went to a no name college like me. Well, I'll list the fact that I interned at Google way above the fact where I went to college. So, you know, it really depends. But I think once you think about that principle, a lot of the other advice really starts to make sense and you can, you know, apply it to your own self.

That's really great. And I love how you approach this question by putting yourself in the shoes of the audience, which are ultimately the recruiters.

And people hear this and they, they're like, yeah, okay, great, but my resume is different. They're going to spend one minute on it. You know, put a timer for 10 seconds, see how fast 10 seconds goes by and realize how little can be understood or read in 10 seconds, you know, so you really got to actually like time yourself for 10 seconds and know what 10 seconds feels like to get a good resume. Because it's really easy to still know all this and then make like a two page resume that has way too much stuff.

Now related to this, in the book, you outline four resume principles to live by if you're applying for jobs in the data science space. I think one of the most difficult steps in landing a job in data science is actually getting interviewed. Personally, I've had dozens of companies reject my resume despite having a graduate degree and work experience at Amazon or a faang. And optimizing your CV is a great way to get past that hurdle. Do you mind expanding over some of your favorite principles?

Yeah, absolutely. Man, it is a tough thing out there. Especially it hurts when people say, oh, the hiring market's so hot, it's so easy. But then Also, when you're on the job hunt and you're getting rejected left and right, even if you've had faang internships or, you know, had some really top notch experience, it's just a grind for everybody. So I've been there. We've been there. That's actually why we wrote the book. Not because we are the world's best job hunters that like landed our first role, but because we ourselves struggled with all of this. So that's a great question. I think our answer for like, our favorite, like my personal favorite resume principle would be the fact that how much license and freedom you have to make it your own and not follow the rules in an effort to meet the fact that you only are going to look at, be looked at for 10 seconds. So now I know that was really complicated. Let me, let me break that down one more time. I'm trying to say, who cares about how wide your margins are, as long as it's readable and as long as you're able to present yourself in 10 seconds, that's cool. Who cares what words you bold? Maybe you want to bold the companies you've worked at because you worked at some name brand companies. Go for it. Or maybe you worked at some random companies, but you have had some really relevant titles like Data Science, then Data Science Manager and then Director of Data Science. And you want to show off that you had this progression, even if that's companies no one's heard of. Go bold that. Or maybe you've had some really good experience where you like save the company 200k, go list that and bold that part, you know, so bolding. It's up to you. Sizes, fonts, it's up to you. I used to make the, this is really hacky. I used to make the fact that I worked at Google that said Google bigger than the other companies and I listed it first. It's just a subtle like two point difference. So even within my own resume, I didn't have it consistently. Like each company was the same font size, but each company wasn't what I wanted you to ask me about. I wanted to ask you. Have you asked me about Google? So I made that a little bit bigger. Sure it looks a little weird, but it gets the point across that this guy worked at Google. So this is what I mean to say. Like, even your fonts don't have to be. Sorry, font sizes don't have to be consistent within a resume section. So that's kind of like my biggest tip is just like, yo, you have a lot More freedom as long as you're in pursuit of this 10 second goal. Not so you can look really cool, not so it can be really visually creative or interesting. You know, some people do those really colorful, well designed resumes. Like, I'm not talking about any of that. I'm just talking about whatever it takes to get read and understood in 10 seconds.

Said another way, you can almost ab test your resume, assuming you apply to enough companies.

Yeah, it's almost always working back from the goal here. You know, if we want to put our data science for marketing hat on, as an applicant, your objective is to increase your conversion rate from application to interview. And all of these subtle techniques are geared towards this goal.

Absolutely. I mean that's, I mean I was on Facebook's growth team and Kevin was a data scientist on Facebook groups day in and day out. That's exactly what we're trying to do. Like, hey, we want to optimize this conversion rate. Let's make the color of this button brighter, let's make it bigger, let's make it bold. I mean, that's the same thing on your resume. Whatever is really good, we make those things bigger. And whatever is neutral, we get rid of it. Or we try to minimize because neutral just bogs down what's good. So that's another big thing. I love to tell people about their resumes. Like people love to put in neutral information. Like for example, in the US I see so many people who list that they know foreign languages like Spanish or Hindi or Chinese. Okay, that's great. But which of these data science jobs needs a foreign language? Nobody, you know, and that's great that you're multilingual, trilingual, but realistically, the more stuff I have to scan past, the harder it is for me to even understand what you do. Right. So, so just so crazy when I see people like list all kinds of random stuff that they think spruces up their resume and makes it look more full, when in fact it actually just detracts from what's good about them. And that neutral actually ends up hurting you. You only realize if you look at it for 10 seconds, you realize, crap, I can't even understand what this person's about in 10 seconds because they put so much random junk and so much neutral information.

I think one common neutral information that you find in a lot of CVs is that header text introducing an applicant that often reads highly motiv, professional seeking position, etc. Etc. I think that takes away a lot from a resume, right?

Yeah, I'm a Hard working, detail oriented data scientist right at the top. Oh yeah. I mean because let's be honest, the first thing you do is where did they go to school or what company did they work at most previously and what was their title? That's what I want to know. So, right. Why is there that like three, four sentence objective right at the top? That's quite Mokato. So once, once you think about it like that way, then you realize how many different things are kind of irrelevant and how many things just like don't matter as much as what traditional advice says. Yeah, so we talk a lot more about that in the book. Like the actual, I think like 30 so specific tips. But once you keep those higher level principles in mind, a lot of it will start to just click.

Now obviously outside of the CV itself, a major part of building an appealing data science profile or resume are projects and building a portfolio of projects. What do you think a good portfolio looks like and what are the principles you recommend here for candidates to stand out?

Yeah, I think we can start actually talking a little bit about what bad projects look like and then you know, Nick can talk about what good projects look like and he has a bunch of sort of great personal examples. I think the clear kind of one liner here is you don't want to use data sets that everyone already knows about and you know, uses and is for a very specific use case. Right. So in school or in classes, online classes, like, you know, everyone's using like the Titanic data set Mnistic. You know, these kinds of data sets are really boring. Everyone knows about them. It doesn't really tell anything about you. You know, through our like career coaching and just, you know, talking with a lot of, you know, candidates and people at these companies, like we've seen a wide range of projects and consistently this kind of stands out. You know, whenever someone says, oh, you know what I want on Kaggle and I did the Titanic competition and I got, you know, like I got accuracy of this much and AUC of this much like that. That doesn't tell us anything about you. Right. It doesn't stand out.

So yeah, so good, good means something that actually tells me about you, that you're interested, that you're passionate, that you actually went out of your way to build something. They didn't just assign it to you or it wasn't just something you had to do in school, you actually really cared enough, maybe you even scraped your own data. We love to see that because it speaks to your own software engineering or like hackiness, which we Love. And I know a lot of companies really like that. So you scrape your own data about something you're passionate about. It tells me, like, this person's an interesting person. And Adele, here's something really else interesting. It's called the halo effect, which is traditionally, if someone is an attractive person, we think of themselves as nicer, smarter. All kinds of positive attributes get attributed if you're a good looking person. Now, something similar actually happens when you're passionate in a job interview. If you're passionate about your own project, suddenly you come across not as passionate about just that, but about data science, about the role, about the company as a person. People just want to be around you. They want to work with you when you're passionate. And here's the thing, you're just passionate for your own project, right? So that's like, easy to get, like, hype about your own project. I'll give you an example. In college, I worked on this thing called rap stock IO. I loved fantasy football. This idea of like betting on football players and how they do in a season and awarding points based on that. And I was like, why can't something like that exist for music? And I love hip hop. I'm a huge Drake fan. Drake's number one fan, certified lover boy to the core. And I made. Yeah, no, it's, it's, it's. It's a sickness. It's. It's like I got a. I got treated.

He's not joking, by the way. Check out his Instagram.

Yeah, you. Yeah, I basically Instagram. My Instagram is basically a Drake fan account where I just, like, pretend I'm Drake. I'm trying to be the data science. People ask me like, nick, what's your next move? And I'm like, I'm just trying to be the Drake of data science. But we'll, you know, have me back on for another podcast where I'll tell you about.

You're definitely on the way.

Yeah, yeah, I'm on the way. No, but anyway, so. So I made this startup around hip hop music called Rapstock IO, where I treated these rappers as like stocks that you could invest in. Go long or short based on how much you believe that their commercial success would be. And I would track their commercial success via Spotify. Okay. So I built this project. I grew to about 2000 monthly active users. And then when I sent cold emails, when I let people at Facebook's growth team know, they're like, oh, wow, this person actually knows how to build things. Likes data, likes experimentation, actually has Built real software. We want to hire him. But here's the crazy thing, right? That's me talking to Facebook's growth team, where I eventually worked. But I use that same kind of story and passion at Uber's growth team and Airbnb's growth team and Snap's growth team. But when I also applied to fintech companies, Adele, I told them all about the stock market, and they loved that aspect of, like, how did you do transactions? How did you do longs and shorts? And when I talked to data companies, they're like, yo, tell me more about the algorithm of how you're pricing these wrappers. Like, tell me more about the data sets you use, how hard it was to use the data. So I'm basically just trying to say, at the end of the day, this one project carried me through so many different companies and made me come across as super passionate. And yeah, I love my data, I love my growth engineering, but at the end of the day, I love my hip hop. And that shines through. And then people remember me as like, oh, yeah, that's the guy with that website. Like, that's super cool. And, you know, for fun, I actually DJ as well. So then the personal side comes up and they're like, oh, I want to work with this guy. Like, yeah, everyone else can code, but this guy actually does something with it and we like him as a person. And I can remember that guy. You interview 10 people, you'll remember that this guy made that weird startup and does DJing and all that stuff. So this is what I'm trying to say, like a well crafted portfolio project, that's something that's not the Titanic data set on Kaggle. Something that's not just like a basic project that really showcases what your professional interest is or what you're about, what kinds of things you like to build. Once you do that, you talk so much more articulately about it, so much more passionately, and it just improves your whole vibe. And in an interview, and it's something that we feel like can really set you apart during the interview process.

It's really about integrating more authenticity into your interview process with your passion and ability to articulate the intersection of your personal interests and your ambition.

Right.

Noah Gift was on the podcast last year and he talked about what makes a great data science portfolio. And he mentioned that it produces thought leadership around a specific topic where you can take a previously unexplored data set and produce an original insight out of it. In a lot of ways, this mirrors what you're discussing since it requires that passion.

Absolutely. And Kaggle, man, they have so many data sets now. I said don't use Titanic data set unless you're super passionate about the movie Titanic or something. But I'm trying to say Kaggle is fine. There's just so many cool things. I love me some Indian food and I saw there's this whole data set of Indian food recipes. Just like, like 10,000 Indian food recipes. I wanted to do like chicken tikka masala analytics. Like, I was like, I had like 86 questions. And you guarantee no one has explored that data set really in depth? Looking at chicken tikka and trying to like do some analysis there. Right. But it tells you something about me and like my passion for food and cooking. Right. So I'm trying to say you don't have to have this wacky fantasy football idea. Like on Kaggle, you like cooking, there's cooking data sets. You like basketball, there's basketball data sets. Like, you got everything right there. And so many questions can be asked and explored. And I think it's just up to you, waiting for you to do that work.

That's awesome. And I should definitely scrape a data set for Lebanese food. Something we've mentioned here, and that is definitely a very common challenge for different types of job applicants in the data space is actually getting your foot in the door and getting an interview. I think this is even more exacerbated when recruiters or companies simply ghost applicants, despite putting in the time and effort to write up a personalized resume or cover letter. In the book, you outline cold emailing and your best practices for cold emailing recruiters to get noticed by them. Do you mind expanding into that? And can. Can you also expand into why there is kind of like this black hole effect of online job applications where resumes aren't noticed?

Absolutely. Yeah. So chapter three, all about cold emails is my favorite chapter. Right. Because it's a very core part of what's led me and Kevin to a lot of career success that we think data scientists, machine learning engineers, data analysts just don't know about. But people in the sales world, marketing world, they know all about this. So cold email is where you write an email to somebody you don't know. So it's like not a warm introduction. It's like just a total stranger. And it doesn't have to be an email. It could be a Twitter DM, it could be a LinkedIn DM. You know, it could be a connection request, whatever you have. It's this idea that you can approach people and pitch them on you. And you don't have to just apply online on LinkedIn or Indeed and be one of the 300 applicants and just wait there to be filtered out by some recruiter who doesn't really know who you are or like why you're a good fit. It's up to you to go pitch the hiring manager, the recruiter, like, hey, I'm Adele. Here's the work I've done. Here's a link. I'm such a good fit for this role for this reason. Let's start the interview process next week. You can send that email tomorrow. Any one of us can send that email tomorrow. We don't have to wait to get filtered to be in front of the hiring manager and pitch them on why you're a good fit. Of course, like, this is not a bulletproof technique. Like, you have to be respond worthy. Like, you have to be relevant. You can't just spray and pray and be random. And that's kind of why you pick your portfolio project in a space that shows your own passion, but also your professional interest. Right? As I said, I'm interested in fintech, growth engineering and data. So I did a project that kind of encompassed each of those things. So when I wrote to these hiring managers about growth engineering, I'm telling them about how I grew to 2000 monthly active users. And when I'm talking to fintech, I'm talking about the stock market aspect and how I'm interested in finance and like consumer finance or consumer product. The point being here, that cold email lets you just really tell a story in a way and control the narrative and like really get in front of people in the way that LinkedIn and indeed just never will let you do because you're just one of 300 applicants. And most of the time you get filtered out. Actually, let's be honest, you don't even get filtered out. You just don't even know what happens. Most of the time. You never even hear back. You don't even get rejected. You don't hear back. At least in an email you can follow up once or twice and we talk about like, what you should actually say in the emails in the book. But I think it's such a good technique. And I should have mentioned one thing. My last job came from sending a cold email. I cold emailed my way to interviews at Airbnb, at Uber, but my, like actual last job came from writing the CEO of the small startup I worked at telling them, like, hey, I love your company. Here's why I'm a good fit. Few days later, we're interviewing. And then soon. I worked there for almost two years. You know, it all came from just an email I sent at like 2am One night night when I was a little bored and thought their company was cool. So I think it's definitely something a lot more people need to do to escape this black hole in the online job portal where you just never hear back.

Yeah, it's really interesting. Someone like me, for example, started off as a data scientist, but now sits kind of at the intersection of marketing and data science. I think now, only now do I realize the importance of being in someone's inbox and being able to reach them and tell them this what I'm all about. And this creates a strong connection down the line.

Exactly. So that's why that's our favorite thing. Because, you know, now that you're doing this podcast and you're reaching out to other hosts and people, you're used to sending these cold emails. But most job seekers, especially in tech, aren't used to it. But what we let people know is even though this sounds foreign, you better believe salespeople, VCs, recruiters, they're sending emails all the time randomly to people, right? Like you can reach out to recruiters randomly, so why can't you yourself reach out to the recruiters if they reach out to you randomly? So it's just something we want to normalize. And we talk in the book, like the exact scripts to send and the real cold emails that I and Kevin have actually sent that have landed us interviews and jobs. And like what we should write, it's all there in chapter three of the book.

So yeah, datacamp's mission is to democratize data skills for everyone, closing data skill gaps and helping make better data driven decisions. Data science and analytics are rapidly shaping every aspect of our lives and our businesses, and we're collecting more data than ever before. But not everyone is able to efficiently analyze all that data to extract meaningful insights. Datacamp upskills companies and individuals on the skills they need to work with data in the real world. Learn more at datacamp.

What's really nice is that it puts the power back in the hands of the candidate. The remaining chapters of the book discuss technical questions data scientists should be able to answer in a data science interview, whether probability and statistics or coding best practices. I think those chapters are gold mines of questions that summarize the technical aspects of data science. Do you mind describing the different sections in those books? More importantly, do you Think that applicants should pay special focus to one type of skill over another, depending on the role they're applying for.

Yeah, sure. So our book again, data science, is super interdisciplinary, right. So that's like one huge thing that, you know, both Nick and I kind of recognize coming from more software background, initially in college. And so we cover the whole range of topics. Right. We cover probability statistics, ML, there's SQL and databases, coding and then you know, product sense and case studies. So really kind of the whole, the whole gamut, you know, there's no perfect answer for what you should pay very special focus to. I think there's kind of two good rules of thumbs that we've come across. So one is to always look at the job description. Right. Like most companies, obviously some companies don't know the kinds of roles that they want to hire for. That's kind of another topic. But you know, generally speaking they've kind of, you know, at least tried to figure out what kind of role they want and try to figure out, you know, hey, here are the technical skills that that person should have. Right. And so if you're for example applying to like an ML engineering role, that is going to have a very distinct and different set of skills than like a data analyst role. Right. And obviously it also always depends on the industry that's in. Right. As well. So you know, broadly speaking, one is look at job descriptions and the second is I think there's a rough spectrum of like, you know, less technical, kind of more product oriented kinds of roles. So for example, at Facebook it was very focused on, I was very focused on sort of product analytics. Right. And so generally those types of roles, kind of like data analyst or product data scientists are going to be much more focused on product thinking about the product like using SQL and basic Python or R versus on the other side of the spectrum, something more like ML engineering, which is very coding and ML heavy or quant roles, which are very quantitatively heavy. There's kind of a spectrum that you can kind of map out. And so that's kind of another good rule of thumb to keep in mind.

Yeah, I just want to add like just ask sometimes your recruiter or the hiring manager, you just send them an email saying, hey, just want to curious, like what does your technical screen cover? And they'll usually be like, oh, we'll talk about SQL in your past projects and then that tells you what to focus on. But I think ultimately practice makes perfect. And that's why our book has 201 like real interview questions from Wall street and FAANG and some of these unicorn startups. I think ultimately this field has so many different types of questions that even if you try to know what to focus on, what one company calls data science, another person might have a different idea of what data science should be at their company. So at some level you just kind of have to know a little bit of everything and practice a little bit of everything.

What I really enjoyed the most was how diverse the set of questions are and how much they cover the data science workflow. In another life, I wanted to be a management consultant and there were a lot of resources on how to crack the case study interview. And this feels like the closest thing I found in terms of completeness in the data space.

Cracking the. There's cracking the case interview and there's case in point or case interview secrets. There's cracking the PM interview, there's cracking the coding interview. We were inspired by all those books because we were just like, why does this not exist? Because these interviews are tough and like to prepare. You could read like 700 different medium articles and read like 5 textbooks and like look at 87 sites or you just read some book. Like that's what these books did. So that's exactly it. Like we tried our best to encapsulate the whole interview experience into one book.

Exactly. One of the last chapters in the book outlines questions around something called Product Sense and really tries to codify the business acumen data scientists need to know and prepares interviewees around that. Right. Arguably this is the most important skill to test for since not necessarily all data scientists are value driven or have this objective function of achieving business objectives. Since this is relatively open ended, I'd love to know the process by which you outline this chapter and if you can summarize some of the key best practices hopeful applicants can adopt here.

Absolutely. Yeah. So let me, let me add a little bit more color. Right. So this is chapter 10 about product sense. Which companies that are hiring for product data scientists, product analysts, marketing analytics jobs, they're going to be asking these kind of questions because it's not just about like building the best regression model, it's about like actually solving the right business questions or like working with stakeholders to figure out like what are the questions we should be answering. So that's why a lot of these companies for these kind of more product oriented roles, product analytics roles will be asking these product sense questions as well as business analytics rules. So I think there's so much variety in this. But what we kind of tried to map against was what we saw very commonly being asset, Facebook, Google and Amazon. Because that represents a lot of different types of jobs. And I think for those kind of questions in the book, we talk about some of the most frequent product sense questions that are actually asked in the interview. But what framework we talk about there that I think, you know, should help for lots of different types of interviews is this idea of, first you got to clarify your answer and make sure your answer aligns to the product and business goal before even giving the real answer. So if there's one thing we've noticed by coaching hundreds of people into these kind of jobs and like actually doing mock interviews, what we've noticed is people just jump in. It's so damn annoying. We try to like have them think critically and instead they just jump into an answer, right? And this is one of those types of places in the interview where it's like, it's not about the answer you give or how fast you came up with it. It's about the types of questions you ask or how well you're able to understand and frame the problem, right? This is Data Frame podcast. Remember that? You got to frame the problem, you got to clarify the problem. It's not about just jumping in and trying to give an answer, right? So I think that's one of the big things that we talk about in the framework is like, first clarify, what are they asking? What are the success metrics? What are we optimizing for? What's the business motivation that's motivating the problem? So let's make it a little concrete, right? If, if the question is, how would you design Uber's surge pricing algorithm, right? You want to clarify like, well, why are we building this? Is it to balance supply and demand? Is it that the drivers have been asking for it? Is it that writers have been asking for it? Is it that, you know, Uber's just trying to maximize their revenue and they see this as a revenue maximization opportunity, or is customer happiness something here? People are just pissed off. Why does it take 50 minutes to get an Uber? I wish you could bring more people on, you know, bring more cars out on the market. Like, we want to ask these kind of questions because it might seem obvious, but like, once you work with someone, it's like, oh, wow, this is a really complicated problem. So I think clarify and aligning is my framework answer, like the first two things. And I think, Kevin, you're big on trade offs, so mention that.

Yeah, absolutely. Just like Nick said before, there's no sort of perfect answer. For a lot of these things. So similarly, there's generally not a single metric that works. Right. And you should always let the interviewer know that you're thinking about various ways to approach the problem. Various metrics, for example. A simple example is there's always going to be countermetrics, right? So as part of Facebook groups, right, like we, you know, sort of the goal of the org was to kind of reduce bad content on and bad actors on Facebook groups. But you know, you can just do that simply by like getting rid of or shutting down most groups. Right. But that obviously hurts engagement. Right. And so it's great to consider, you know, suites of countermetrics as well. And that's something that we see candidates not doing enough of. And then outside of that, you know, again, there's no silver bullet. So, you know, product intuition alone or like just because the a B test is, you know, says to do so, like, doesn't necessarily mean immediately that you should like build something or ship something. Right? So there's like a lot of real world kind of like cost benefit analysis to think about. So, you know, in general, like trade offs is always, always super, super important. And another thing to keep in mind, like, is that, you know, these interviewers, generally speaking now, it's not always true, but a lot of times if they're asking about like products in their domain expertise, like they thought about a lot of these problems like for much longer and way harder than any applicants have. Right? And so that's the whole reason why they're trying to gauge your intuition and the questions you kind of ask, you know, something that's kind of like new to you and you know, they're really just trying to assess like how you think. Right. And I think that's why kind of to Nick's point, like it's most important to just really kind of think critically and ask a lot of questions and you know, hopefully the interview should be fine.

Circling back maybe here to the case study interview in management consulting, you know, a big aspect of that case study interview is not necessarily having the right answer, but being able to clearly articulate sound thinking when solving the case. That I think is the biggest win when it comes to product sense, right?

Absolutely. And that's why these frameworks are so important and we talk about them in the book and put real questions in there with the real solutions. Because even if we talk about this framework from coaching so many people, we tell you all this stuff like clarify, align to the product and business goal and then mention trade offs and I Hit you with that problem, like, hey, what are some success metrics you use for Facebook dating? And we'll just see people just jump in. Oh yeah, I'd use this metric. I'm like, well, what happened to clarifying, like, what is Facebook dating trying to do? What's the, you know, it's so easy to just like forget about the framework, which is why practicing makes perfect, because it's so easy to just like go off on your own.

Exactly. And I'd love to pivot here, maybe to discuss the hiring manager perspective instead. Now, of course, I'm sure preparing for this book meant that you've also spoken with a lot of hiring managers who've been hiring data scientists. This I'd love to know, given your close work with them. What do you think are some of their best practices hiring managers need to adopt when hiring talent? And what are some of the biggest pitfalls they should avoid?

Yeah, absolutely. Yeah. So in addition to just talking to a lot of recruiters, hiring managers and VPs, Kevin and I have also hired some people in our own past. So we've seen it both as a practitioner and as like someone being interviewed and then also talking to other people. So it's something that's near and dear to our heart because ultimately we're thinking like, hey, not just how can people interview better, but like, how can people get better talent? Like, that's something top of mind for us as well. So I think one very interesting thing that there's a lot of debate around, which I want to stoke again here, is whether take home challenges are a good thing or not. So I just want people to realize that a lot of senior talent or in demand talent will not do your six hour take home challenge that you think takes six hours, but actually takes like, like 12 hours, you know, or like a two hour thing. Sometimes just setting up for a two hour project takes like 45 minutes because it's just building context around it. You know, people just don't realize that. So I think take home challenges you need, people need to be really thoughtful that about their time limit for a take home challenge and realize that they might be doing some adverse selection where it's like, hey, the best candidates might not do the take home. So I think that's one thing that we want hiring managers to really intentionally think about. Another interesting thing is speed really matters, especially at smaller companies. Okay. So in the sales and marketing world, we know that time kills deals. It's all about speed and you want to close a deal fast. Hiring is a Lot like that. Okay, and here's the thing. Facebook and Google, they take a long time to hire their candidates. Like I'll give you an example. Google, they have like committees on committees, a hiring manager committee, a compensation committee. And you know what, people might put up with it because they're Google. But there's enough talent that just like, hey, I don't want to wait two and a half months to hear if I have the job at Google. I'm trying to job hunt next month, or I already have two or three offers in hand. Why am I going to wait an extra month and a half for Google when I have two or three decent ones? So I think another thing we let hiring managers know is if you're not Google, you can't take two months and be wishy washy. You have to be decisive and communicate well because you can't hide behind, oh, we had a 16 person committee to decide your offer when your company's only 16 people. So I think speed matters and use that to your advantage. And the other thing I want to bring up is the primacy effect. So it's where whatever you know, first, we tend to like or weigh more. So it's a real, real thing that hiring managers can use to their advantage. So what happens is before you go on the job hunt as a candidate, you're thinking, yeah, I want to maximize my compensation, I want to get, maximize my offers. I'm going to interview with 10 people and try to get six offers and play them all off each other. But guess what actually happens? The first company that gives you a pretty decent offer, you're like, oh, I like this company, they like me. You lose a little steam to keep interviewing after that because you're just kind of couching like, ah, do I really like this company as much as the first company? I already have one offer. I'm getting a little tired with these technical interviews. So I think there's a real advantage to being the first one to give someone an offer. And again, that's where the time plays into it because tenants want to anchor like, oh yeah, this is a pretty decent offer. Like I don't know if it's worth shopping around. And like hiring managers can use that to their advantage. Yeah. And then the other last thing is just like selling people on what actually is very unique about your company. And I think this takes a lot of self reflection from a hiring manager to even answer. Like every company says, oh, we like to have work hard and have fun or we like to do this or that. But I think it takes a lot of humility to be like, hey guys, our company is pretty chill. And I'm gonna tell you that straight up, like, this is a very good company for work life balance. And here's the other thing. If your company is intense, that's also okay. You can say that. But then my last job, they said, hey, this is a high hours role. This is not. This is a small startup with a high hours role. We expect a lot of hours. And guess what? Two thirds of people are like, nah, I don't want to interview here. But one third of the people who are crazy enough to interview, like, interviewed there. And I think people want to try to appeal to everybody. And when you think about in marketing and positioning, if you try to appeal to everybody, you appeal to nobody in particular. And in this crazy market where each company is doing something unique and standing out, you can't get away with just trying to appeal little bit to everybody. You got to be a little bit more unique and know that. So I think it's very important for your own company and your leadership to have sound positioning on why is this company unique and what's something special. We do, do we have really good work life balance or really bad, but give you a lot of growth opportunities? Do we pay you a lot of money or do we pay you not a lot of money and be upfront that this is not a lot of money, but we're going to invest in people and like really train them because they're undervalued and we want to invest in you. You know, that kind of humility and like candor is really refreshing because it stands out against the sea of other companies that are all just doing the same thing. So that's my tips for hiring managers to like really get good talent.

That's really great and I definitely agree on the honesty aspect of it as well and letting candidates know what they're getting into. I think in our conversation so far, it's been clear that applicants need to think like marketers and they need to creatively think about how to get noticed. Similarly, there are a lot of data team and hiring managers that need to think about ways to attract talent and compete with the fangs of the world. Right. What are ways data teams can think like marketers to attract talent?

Yeah, absolutely. So I think one thing is we love companies that have good engineering blogs or data science blogs because it gives candidates something to latch onto, like, oh, this is the kind of work they do and it lets your own team look good. And I think Ultimately, people want to work with other people. People don't want to work at this, like, nameless brand or company. They want to work with Joe or Bob or Sally, you know, and having these kind of technical blogs authored with like, hey, at the bottom, like, a call to action. If you like this blog and you love thinking about transportation, come join our company and work with Joe. Joe previously worked here, here, here, and he loves solving this thing, like, humanizing that person, because essentially, your engineering blog is a really great way to attract talent. So I think just putting that call to action, you know, in marketing, we call it CTA Call to action, right at the bottom of, like, hey, like, I want to work with this author and make it really easy. Like, hey, here's a link to the careers. If you. If you like this guy and you like this person's blog, let's do that. So I think just putting more call to actions in your materials and actually just showcasing your own company's unique values. I think that's. That's something big. And I think, again, going back to the marketing thing and positioning, nailing your positioning is very important. And I think it's really up to you as a hiring manager to work with your leadership team or CEO to really understand what makes your company unique. And if nothing makes your company unique, you know, I mean, I think every company is doing something interesting or different, like, you know, because otherwise, how could it compete?

Right?

There's something unique about each company out there. Otherwise it gets squashed by competition. I don't know too much econ. That's more Kevin's. Kevin's Avenue. But I. You can't just be doing what everyone else is doing. And I think that it's up to you to tell that story effectively. So I think that's another thing. It's like figuring out what's unique and positioning that, showing that off at every stage of the interview. And ultimately, I just want to say this one last piece, which is this. This whole hiring talent thing is about how do you make a candidate feel valued and special at scale? And I know that seems like a contradiction. You want to make them feel individual, special and unique, except at scale. How do you do that? Right? So once you frame the problem like that, that gives you really good ideas for, hey, how do we up our scale? And what technology systems or process can we do upper scale or what can we add to make you feel even more unique so that we write more personalized emails, we send you a personalized gift, we send you company swag after you interview with us, we send you, you know, some. We give you a free trial of the product in the beginning of the interview so that you really get to sense, like, what our company offers. We give you, you know, if it's aws, let's give you some AWS credits. You know, what if they hit me up with that? I mean, aws, Amazon, they're big, so they might not need to do that. But, like, if you're a developer tool or data science tool, you can. You can offer all those things you can send your. The CEO is a big believer. Like, let's say this company does some very generic stuff. Stuff. But they're a big believer in, like, the lean startup and the lean movement and like, hey, being like, very lean and efficient. If you're interviewing a candidate, why can't you just send them the book for free, the lean startup, and send them that lean production book, the Toyota way, right? So that's like, hey, like, this is what we believe in in our company. We don't pay that well. We're very efficient and we're. But we're very systematic and we do a lot with less, and we believe in this lean approach, and we want you to join the team, and this is how we think. Boom, you stand out, you know, even if you're paying less and you're a little bit more of a bootstrap company and all it cost you was two books. You know, that's like 30, 40 bucks easy.

It's great as well. And you kind of give pointers based on company size, how to fund these activities. You know, I've been with companies that fly you out and do all these fancy bells and whistles, but there are ways that you can compete with that, even as a lean startup.

Send me a book. Exactly. It doesn't have to be this big thing. And let's be honest, like. Like spending an hour interviewing with a data scientist that costs the company real money. So why. Why try to save some money and not send that $50 gift, $60 gift, when, you know, the whole interview process, hours of a data scientist time to evaluate you cost the company, like, hundreds to thousands of dollars of lost time, wages and things like that and like, focus.

So, yeah, two other things to add there would be, I guess one is if. If you can demonstrate how data driven the firm, like the, you know, the culture is. And just like the firm actually uses data, that's super helpful. The same way that, you know, a lot of engineers, when they're looking on their job hunt, they Want to know, like what would their impact tangibly be, right? So are they building like the product that customers are using? Are they more focusing on like internal tooling? Like what, what are they kind of actually working on? And I think the same way, you know, it's less probably spoken about in public, but, but you know, a lot of firms are trying to, you know, there's a broad spectrum, for example, right, early adopters to kind of more mainstream adoption of data and it's, and its use in firms. But a lot of firms these days, you know, finance, tech, wherever, are trying to become more data driven, right? And so really being able to, you know, demonstrate that, hey, like data plays, for example, at Facebook, right? Everyone knows that like a B tests are so core and experimentation is like so core to the company culture, right? It's definitely like a very attractive kind of selling point. And so that would be kind of like the first additional tip. And then I think the second one is also, we kind of touched upon this earlier but just basically having like good honest job descriptions, right? So there's that phrase, might be butchering it, but you know, it's like happiness is, is the delta between expectations and reality, right? So in the same way, you know, a lot of candidates, especially junior ones, you know, they might have these expectations like, oh, like, you know, I'm going to join this company and I'm going to do like, I'm going to build these ML models that will get this much revenue uplift. And in reality it's like, well, hey everyone, there's a lot of reasons why that probably wouldn't happen for any company in the beginning. And they kind of come in doing some internal tooling or dashboards or something and they kind of are like, oh, this is not the role that I wanted, right? And so I think really kind of making sure that you have like, hey, this is like, you know, this is the, this is the job that you'll be doing, right? And listen, like we want to especially you know, catering toward the audience, right? Like for younger folks. Younger folks are always like, oh, like, you know, especially these days, right? Like at Penn everyone was super. So I went to Penn, right? Everyone's super like career oriented, right? Like, oh, like what's next? And you know, how do I kind of like climb up the ladder the same way? Like, you know, hey, just for the younger candidates, like on the job descriptions or maybe when, when they join your firm, like, just make sure that you're willing to, you know, talk with them and just, just, you know, hey, here's how, here's how you can have more and more impact at the firm and where do you want to go, right? So it's also about, you know, again, it's such a hot job market these days, right? It's not just about, hey, I want to work for you. It's also about like, hey, how can you grow the candidate's career as well, right?

It's this kind of crazy thought process, but like a little bit like, hey, what do you after this job? What are you trying to do? And let's get you to that spot. You know, it's this kind of humility thing because sometimes you're like, oh, this is the last job you'll ever have. And that's like just not a reality, right? So it's a really good thing for if a company can be upfront like that. And this is, we're talking about, like, if you can't compete with Facebook, Google, maybe Facebook doesn't have to be like that and say, hey, come to Facebook. Be here forever, you know, but for a lot of companies, they have to realize like, hey, talent comes and comes and goes. If we can just position why this is such a good opportunity for you right now to get where you want to go and we're aligned to that, that kind of realness. Oh, man, that's so awesome. And I had that in my last job where I said, hey, I want to be an entrepreneur straight up. They asked me, what do you want to do for five years from now? I'm like, hey, if I won't be working for you guys, I'll be running my own company. And the CEO said, great. Where a small startup, we're scrappy, will teach you what you need to do and you're going to build this company right now for the next few years so that you can go do your own company and we support you that and when you do, we're going to write you a check to do that. I said, wow, you're the only company who said you're going to write me a check when I quit. Like, you want me to quit in a few years. I mean, they didn't go that far to say like, I want you to quit, but like that kind of like candor of like, like, hey, we get it. And that's the truth. Most of these early stage startups, people join them because they want to do maybe something entrepreneurial or learn something more, but most companies will pretend like that's not the goal and it's like, oh, this is your Forever home. And talent comes and goes. So just having more honesty in all these conversations and making sure people are aligned to what you're offering always just helps smooth things over.

I definitely echo that. Even at data camp, we especially celebrate team members who exit to become founders of their own. Definitely see where you're coming from. Given we're talking about how different organizations can compete with major tech companies in hiring. Where do you view the role of upskilling when filling out a pipeline of candidates? You know, given the fierce competition over talent, from a hiring manager's perspective, do you think there is a room to hire and upskill as opposed to wait for that unicorn data scientist to join your team?

So we think, you know, a very simple way to put it is, and there's a bit of nuance to it, but, you know, if you have unicorn salary, like, you can get unicorn talent.

Right?

Again, it's a free market supply and demand. If you want to pay for those unicorn data scientists, like, you'll have to meet the market where it's at. Right? It's really simple. That being said, you know, we, we do think that there is a place for upskilling. So as an example, you know, we were talking about, hey, maybe you have a younger candidate or a younger candidate who's like, just very hungry to learn a lot and just, just have more and more impact, right? So we always recommend again, in general that hiring managers just try to learn and kind of try to like map out, hey, what does this person want out of their career? You know, what have they been learning and what do they want to be learning, right? And if you can kind of like make that mental connection that like, hey, this person is, you know, really smart, really hungry, just like, wants to learn a lot. You, you know, we think that it's, it's worth kind of, you know, giving that, giving them a shot at that.

Right? Yeah. In, in startup land, they call this like slope over intercept. It's not about where you are today or where you started. It's about how fast you're growing the slope of your learning curve. And I think that's something that people and data will intuitively know, hiring managers intuitively know. And then you're faced with six resumes and then you just pick the most, like, risk averse choice, you know, and then complain, oh, why do they want so much money? They're perfect on paper and then they're asking for double the salary. You know, people intuitively know this and then they forget about it when faced with a reality and I think so much of this is just like having that humility to be like, hey, unless you're giving that unicorn salary, you're gonna upskilling is very much a real thing. And I think that's like, okay. And it should be celebrated because listen, so many of us data scientists are self taught or even if not even if we have a degree, let's be honest, not all of your professors are amazing. Like there was a lot of late nights grinding, learning, coding, it's a very individual way to learn. Like, like ultimately, you know, most people don't learn by watching someone code, it's by coding themselves. Right? So it's just sort of like if that's our field, can we, can we as hiring managers really embody that and have the courage to like when faced with these resumes, pick that, you know, it's just like a courage thing. And I know it's like not easy. And then, you know, we say all this and then you just look at five resumes and you just pick whichever one's risk averse. But I think it's just ultimately having that humility to realize like, hey, I'm sort of self taught or I'm sort of from a diverse background, so why shouldn't I give this person a chance? Or like, you know, so I think the market pushes people anywhere anyways that way towards being realistic. But I think just like if you can just right from the get go, be realistic and like be making more off, be making offers more intelligently to people who display that growth potential rather than someone who just checks all these random boxes, you're going to have a much smoother time hiring candidates.

Now as we close out, I'd love to pivot to discuss more of the future and how you believe the data science workflow and skillset will change. What do you think are some of the major trends that will shape the role into the next few years? You saw over the past year large language models like codecs, GPT3, AutoML. How do you think this will impact how a successful data scientist or data science applicant is perceived in general?

Yeah, that's a good question. I think the short answer is that again, the rise of a lot of more black box models will kind of only accentuate the need for data scientists. Like there's a, there's a great analogy I think back in, you know, well, I guess a long time ago now, but the ATM was invented, right? And they thought bank tellers would go out of business. And it turns out afterwards, empirically there are actually A lot more bank tellers. So in the same way, if we kind of like, if we kind of just look at the way technology is progressing, right? Like there's been a crazy amount of innovation in the last 20 years. You know, people are talking nowadays, you know, about, about like generalizable AI models and just kind of like super, super ML, if you will. I think that there really will be a kind of like blend. Like there's a, there's a stark example, you know, in the finance industry, right, where there's always this question of like man versus robot, right? And there's like, is man alone the best? Is robot alone the best? Or is like man plus robot is the best, right? And you know, there's obviously like a lot of, you know, up for debate, but generally speaking, there are things that humans are good at that, that machines can't do and there are things vice versa, right? And one of the things that humans are, are great at that machines. So machines are extremely amazing implementation, right? At conducting latency, just any actual implementation, you know, running these, the actual algorithms running, right? And they can run a repetitive process, you know, 10 million times a second, right? That's what they're best at. But humans are, are good at the strategic level of thinking, right? So the same way that, let's say you were to look at Wall street and look at the different kinds of firms, right? So you have like fundamental finance firms where it's very human driven, right? They don't really rely on automation. You have on the other spectrum, quant firms where it's all algorithms kind of running and trading money at the same time. Even at these quant firms, it's not like, I mean, they hire researchers for a reason, right? There's the researchers are there to tell the algorithms how to think essentially, right? And so I think even more and more as, as the tools become more advanced, it basically automates away all of the parts that were frustrating, right? So every data scientist knows there's like a, a de facto kind of workflow, right? So you kind of have to get a bunch of data from somewhere. You have to like analyze, you know, you have to look through it, look, clean it up, do a bunch of like feature exploratory, feature analysis, right? Like you have to do a bunch of stuff and then kind of run a bunch of different models, do some hyper parameter tuning. When that kind of all gets packaged up and simplified, you can now run, instead of running like a model a day or something, you can run like hundreds in like a day, right? And Then that really lets you as kind of like the human architect kind of think about hey, strategically like what, what should, like what kind of business value should I be using? Right. Instead of like doing all the kind of like data munging kind of work. And so I think it will really like, it seems kind of scary. Oh like you know again, GPT3 is amazing. I don't think it will displace all data science jobs. It will kind of just, it will just make data scientists kind of focus more on the strategic kind of like higher level decision making principles.

Completely agree that there will be no data scientists automation problem anytime soon. If anything, data scientists workflow will be supercharged given that automated machine learning or AutoML is increasing and more and more. So there are out of the box solutions that can do a good job. What do you think are going to be the hallmarks of a great data science portfolio?

Yeah, so definitely, you know, the same principles as before that, that Nick talked about apply, you know, obviously like tying in personal interests, being creative in your approach, right. You don't just want to take you know, the kind of the simplest, you know, data set and, and just you know, run like, you know, have GPT3 or some other model run on it and just like say hey, I just ran this and there you go. Right. I think that and as we've seen actually empirically with GPT3 there's been a lot of, for example, even startups born out of, of sitting on top of GPT3. Right. So I think in the same way, you know, especially for those that are in the ML space, right, like just being creative with the approach and you know, obviously trying out different kinds of models and just kind of like being more exploratory and not saying that hey, like the primary value add of, of, of this project in this portfolio is like the actual like model running because that's, that's, that's going to be all automated away, right? But in more of like hey, this is what I'm exploring. This is maybe the sets of models I'm writing together in this particular space or using whatever data sets that's going to be more the strategic again higher level thinking of how does it solve a problem rather than oh hey, I ran a model and here's what it outputted. No one cares about that anymore in the future with AutoML and all these other innovations.

Finally, Nick, Kevin, where can listeners learn more about what you're working on?

Absolutely. So of course you can check out our book Ace the Data Science interview on Amazon it's the number one bestseller and it's the number. Yeah, it's. It's doing good. So y' all should check it out. You can also follow me on LinkedIn. I have 65,000 followers and I post every single day about tech, career advice and data science. So that's just Nick Singh on LinkedIn. You can also check out my own website, nicksing.com where I talk about cold emailing, salary negotiation advice, and other blog posts for technical people to advance in their careers. Kevin.

Yeah, I'm on LinkedIn. Is Kevin Huo. And then I think Nick forgot to mention, if you want to see the Drake side of things, his. His Instagram handle.

I was about to mention that.

Yeah, true. You can follow us on Instagram. We have a Ace, the Data Science interview Instagram, where we post interview questions and some videos from our talks and different snippets as well as some photos of me looking like Drake or trying to be cool. Yeah, it's not very cool, but I have fun with it, so it is what it is.

Awesome. Nick, Kevin, thanks for coming on the podcast.

Yeah, thanks for having us. This was a lot of fun, Adele. I love what you guys are doing.

Thanks for having us, Adele. This is super fun.

You've been listening to Data Framed, a podcast by Data Camp. Keep connected with us by subscribing to the show in your favorite podcast player. Please give us a free rating, leave a comment and share episodes you love. That helps us keep delivering insights into all things data. Thanks for listening. Until next time.