MLOps Zoomcamp: Free MLOps course. Register here!

DataTalks.Club

Home Articles Slack Events Podcast Books

Moving from Academia to Industry

Season 6, episode 6 of the DataTalks.Club podcast with CJ Jenkins

Links:

Did you like this episode? Check other episodes of the podcast, and register for new events.

Transcript

Alexey: This week we'll talk about moving from academia to industry. We have a special guest today, CJ. CJ worked as a postdoctoral researcher at Martin Luther University, which is a university in Germany and then she decided to move into data science. Since 2018, she has been working as a data scientist and now she is a data science lead. Welcome, CJ. Let's start with your background. Can you tell us about your career journey so far? (58.0)

CJ’s background

CJ: Yeah. I was an evolutionary biologist for a long time. I did my undergraduate Master's and PhD in evolutionary biology. Then, about six months before I finished my PhD, I was thinking “I'm not sure I want to stay in academia.” But that’s not the time to make life-changing decisions, so I just put my head down, finished up, and I was like, “I can figure this out during my postdoc.” Then about a year and a half into a three year postdoc, I was like, “I don't think I want to do this anymore.” So I shifted gears, studied a bunch, and got an awesome job in industry in Berlin. I've been a data scientist ever since. (1:28)

Evolutionary biology

Alexey: What’s evolutionary biology? (2:08)

CJ: Very good question. It looks at how populations of organisms evolve over time. When people see ‘evolutionary biology’ they're like, “And then you ended up in technology?” and I’m like, “Hear me out.” Because evolution happens to a population instead of an individual, so it's rooted a lot in statistics. I think that was one of my edges in getting my first data science job, is that I knew more about statistics than anybody else in the room. I've been building experiments, running statistics on the data and collecting it. I was teaching my students how to do statistics, so – teaching statistics within the courses. So I had a really solid foundation. But on top of that, because evolution looks at how populations change over time, it's really rooted solidly in math and differential equations, and looking at how population dynamics can change. So it's literally just looking at how populations evolve. But it gave me a really strong background for things that I ended up using in data science. (2:11)

Alexey: To prepare for this interview, I was going through your LinkedIn profile. I actually got the impression that you were also doing data science stuff back when you were in academia. (3:16)

CJ: Depends on how you define ‘data science stuff’, because I think that's an interesting question. (3:33)

Alexey: I have a quote. It says you were doing these things: “manipulating big datasets, analysis of A/B tests, working in environments of large and complex data structures, commonly applied advanced statistical and machine learning techniques.” Yeah, so this sounds like a usual data science job. (3:36)

CJ: Yeah. Absolutely. You know, it's funny, because I still rely on a lot of those tools. A lot of genomic data is unstructured text files full of sequence reads that are three to four gigabytes each. Figuring out how to process those in bash requires a certain programming knowledge. But the things that I was missing were the deployment component. I had to Google what an API was when I first started as a data scientist, and I had to figure out what an infrastructure was, what a Docker container was, and all those things. So I had the theoretical understanding, and the backing, and a lot of the coding done. But you can ask anybody at my first job – during my first three months, I just like walked around with this look of terror like, “I have no idea what I'm doing!” [laughs] But I figured it out. Yeah. (4:00)

Learning machine learning

Alexey: Did you need to learn any machine learning or you basically knew everything you needed? (4:45)

CJ: Oh, no. I had to learn a lot of machine learning. Most of the background I had in machine learning was very much based on statistical models. I took my first machine learning course, and everybody was talking about how it’s the best, like “Machine learning! Machine learning!” And then I looked at it and I was like, “The first ones you guys talk about is linear regression and logistic regression. Those are just statistical methods that I've been using for years.” Actually, my first case study for my first job was – we were supposed to build a predictive model so that we could predict transactions of individuals. And I sort of cheated because the data was too sparse. So I just made a proof of concept by combining the data and then predicting the mean. But I use a generalized linear model. It was just a GLM that was tied to a Poisson distribution, I think. I can't remember the exact distribution I used, but it was just a statistical model. (4:50)

Alexey: So basically, for you, the theoretical part of data science, or the theoretical part of machine learning, wasn’t scary. You could just watch a course and then apply everything you learned and quickly grasp all these things. In that sense, the background you had as a researcher helped you a lot. (5:49)

CJ: Exactly. Yeah. (6:08)

Learning on the job and being honest with what you don’t know

Alexey: Then for the things you didn't know – you mentioned you didn't know anything about deployment, you didn't know what API is, Docker, and all this stuff. Yep. Did you have to learn this on the job? It wasn’t before getting the job, right? (6:10)

CJ: No. I feel like I got really lucky. It's one of those things, like I said, and I get asked this a lot from academics, “How do I move into industry?” The unfortunate truth is that I got really lucky with time and place. I moved there at a good time, whereas I think the field is really competitive now. They also talked about this a lot at N26, which was my first job. They came back from my interview and just looked at everybody else in the team and they were like, “She said everything you're not supposed to say in an interview.” But I was very open with them. I was like, “Here's a list of things that I'm bad at. But I really want to learn from you guys. I don't know everything, but I want to learn.” (6:25)

CJ: But one of those was that I didn't know Python when I started. I had to learn that on the job. If I'm being totally honest, I didn't know exactly what a position in data science was going to look like. My sister was asking me ahead of time, “So, what are you going to do?” And I was like, “You know, manipulate data… use data to gain insights. Make data into value.” And she was like, “Yeah, none of those mean anything.” And I was like “I know.” Until I started the job, I wasn't sure what that was gonna be. It was a hard crash course. But I think there were a couple things that helped me succeed and a couple of them came from academia. (6:25)

CJ: One thing in academia is that you need to be able to teach yourself stuff. At that point, I had 14 years experience teaching myself how to do things and teaching myself how to learn things. So I did a lot of that. I think one of my colleagues is listening, so they can attest to this – everybody else at the company started at 10. The core hours were 10 to 6. And I was usually in the office between 6 and 7 and I spent those hours – between whenever I got in and 10, 10:30, when our team did our stand up – just learning new things. Everyone was like, “You know that you don't need to be in here this early?” And I was like, “You know I don't actually know what I'm doing yet, right?” I spent a lot of time figuring it out back then. (6:25)

CJ: I also had incredible colleagues who were willing to sit down and do a lot of pair programming with me, which is still my favorite tool with new people – just to sit down and have two people working together to solve the problem. Then I could see how to solve the problems by myself. But yeah – I didn't know any of that when I started. (6:25)

Convincing that you will be useful

Alexey: That's cool. I'm just trying to think – some people are listening to this and they're thinking, “Okay, now I also have this theoretical foundation. I know statistics well. I know machine learning well. But how can I convince a potential employer that they should hire me and then I learn everything on the job?” Let's say, if you don't know Python, if you don't know deployment, you don't know API? All you know well is statistics and machine learning. So how do you convince an employer to hire you? How do you convince them that you will learn everything in, let's say, three months, and you'll be able to start being useful? (8:41)

CJ: That's a very good question. Again, I think I got really lucky. But from the other side now. Now I'm leading and building a data team, so I can speak to what I look for in people who might not have a lot of experience. I am specifically going to be hiring two new junior people in the spring. But what I'm looking for is three things. That they're reasonably smart – they can pick up on concepts quickly. That they have the ambition to learn – they're open to learning and trying new things. And that they can take feedback. (9:25)

CJ: If you're not open to learning new things and you're stuck in your way, it's going to be a problem. If you get really defensive when I say something like, “Hey, maybe you should do this differently.” Then it's going to make it hard. But if you have those three things, I don't care what you know. I can teach you anything. But if you have those three things – those are the kind of people that you want to invest in as a junior. (9:25)

CJ: So from the counterpoint, if it was somebody who's like, “How can I get my foot in the door?” I think demonstrating those things will work. I don't need to see your crazy CV or crazy portfolio, especially for a junior position or an entry position. I want to see that you have the ability to pick things up quickly and that you can take feedback. (9:25)

Alexey: How do you test for these things during the interview? (10:42)

CJ: I'm brutal. [laughs] No, I'm actually really nice. I think this is one of the things that you think a lot about as an academic like, “How do you define smart?” This is one of those skills that transfer really well. It’s somebody's ability to absorb and synthesize new information – they can figure out what I'm telling them and then synthesize it with what they already know. You can just ask questions like, “Tell me about a problem that you've worked on.” And then I'll start throwing curveballs at them and be like, “Oh, do you mean like this?” And then I see how they respond to that. (10:46)

CJ: Then, to see if they can take feedback, I just flat out ask them, “Tell me about a time when you were wrong.” I think that's a very different question. I had somebody answer this the other day. They told me about a time that they had failed. I was like, “That's not what I was asking. Tell me about a time when you were wrong.” I can give you a list of the ways that I'm wrong. I'm gonna run a whole series at one of the universities in Stockholm about ways that machine learning algorithms in production fail. But being able to admit, “Hey, I was wrong about this, and this is what I wanted to learn from that. Here's how it changed me in the future.” That means that you're open to the possibility that you can be wrong, and you're open to other people telling you that. I think that's important. (10:46)

CJ’s first interview

Alexey: Since you mentioned that you will be hiring juniors soon. I guess many people are taking notes right now. [laughs] Yeah, but that's tricky to check for these three things. But I guess companies also check their theoretical background and how well you know all these things. Do you remember how the interview actually looked like for you? Like when you were transitioning – the interview with N26. (11:59)

CJ: Yeah, I remember it very distinctly. I have a good memory. But this one stands out in my mind because it was such a weird experience. What was the interview process like?  Okay, it was really easy. [laughs] I got so lucky. I had a friend in Berlin who had gone on a Tinder date with a guy at N26 and she was still in contact with him. So she told me, “CJ, do you need a hook-up?” I applied to four positions at the time. Again, luck. I applied to four positions, I had one interview, and I got the job. And it was my number one choice. This kind of thing doesn't happen anymore. This is not something that happens these days. I got lucky. (12:34)

CJ: At N26, I skipped the recruiting step, because I was a referral. They sent me a case study and I spent like a solid week working really hard on it. At the time, all of the deployments were in Python. But at the time, they said, “You can do this in Python or R. We don't care about the programming language. We just want to see that you know the concepts.” So I did it in R because that was my native language. I sent them this case study. A week later, they contacted me, “We’d like to interview you.” I was like, “Okay. Cool.”[swooning] So I sat down with the two data scientists at N26 at the time and, like I said, it was an hour and a half interview. (12:34)

CJ: I was just like, “Here are all the things that I don't know. I'm really excited to learn from you guys.” So all the things that you're not supposed to say in an interview. But I think they appreciated the honesty and I felt more confident from that point forward, because I was like, “They know exactly what they're getting. I don't have to have imposter syndrome. They know my failings before I get in the door. I don't have to pretend to be something.” (12:34)

CJ: Everybody else was like, “Did you fake it till you made it?” I was like, “No, absolutely not.” Then after that, it was two weeks and then I had an interview with the CDO at the time – chief data officer. I'd studied Python in between. I was like, “I have to know Python terms!” He, at one point, literally pulled out a computer and was like, “Walk me through what this Python code is doing.” I could pick up logically what was happening, but I didn't know any of the syntax. So we talked about that and then, two days later I had a job offer. (12:34)

Alexey: Yeah, that's pretty impressive. (14:59)

CJ: Yeah, it was very easy. It was a lot of luck. But I walked out of the first interview – and keep in mind, up to this point, I've had 14 years as an academic, so I've spent my whole life in labs and universities – and I walked out of this interview in this super-cool tech startup. Everyone was walking around looking like cool startup tech people in the office. I got to the sidewalk and I just bent over laughing so hard, like “I can't believe this is my life right now. I can't believe that this is where I am.” But it was time to leave academia. So I guess we're moving forward. Yeah. It was bizarre. (15:01)

Transitioning to industry

Alexey: Did it take a lot of time between when the idea occurred to you that you wanted to try something else to your actual first day at N26? (15:36)

CJ: It was about a year – a little over a year. I was in London – which is one of my favorite cities in the world – visiting a friend in January. It hit me all of a sudden. It had been building up for a while. It was like ‘death by 1000 cuts’ but that was the final cut where I was just like, “I want to be able to choose to live in a city like this someday. If I stay in academia, that's never going to be possible. I'm not gonna be able to make choices about my life. I'm not gonna be able to choose where I want to live. I’m never gonna have enough money to do things like this. So it's time to go.” But I looked at my schedule and at that point, in January, I had committed to a number of things. (15:47)

CJ: I committed to teaching that spring, I committed to field season that summer, I had a couple of PhD students and Master’s students that I needed to get into a good position before I left them. And I had a huge talk that I was invited to give at this conference at the end of August. So I was just like, “Okay, I'm gonna shelve that idea until August.” I just worked really hard as an academic knowing I was going to leave, but continued working really hard until the end of August. Then I came back from the conference and completely shifted gears. I rewrote my CV, did a Coursera specialization in data science to get a more broad view of it. Then I started applying to companies in Berlin in the middle of November. I started at N26 in the beginning of February. (15:47)

Tailoring your CV

Alexey: Can you tell us a bit about rewriting your CV? Because I told you that, when I was reading your LinkedIn – I don't know how many iterations you did there – but it sounded like very data science related stuff. So I could see that you were already doing all this data science stuff before you were officially doing it. Yeah, I guess this is more like a framing question – how do you put all these 14 years of experience in such a way that people want to hire you? (17:14)

CJ: Yeah. It was hard. And again, I'm super lucky. I had really good friends who were willing to help me. And because I had really good friends who were willing to help me and who had been in industry for a while, I am now always, (people are gonna contact me now) but I'm always willing to help other people. Random people ping me on LinkedIn and are like, “Hey!” And I'm just like, “Yeah, I'll look at your CV. Yeah, we can talk through the interview. Yeah, of course. I'm happy to do it.” I'm pretty active on the DataTalks career channel, because I had help – I couldn’t do this by myself. So I feel like I owe the world help in return, because I got really lucky. (17:47)

CJ: So back to the CV, I was like, “Okay, how do we write my CV? It's like six pages long. I know nobody's gonna read that in industry.” My friend Penny and I did a weekend trip to Italy and she was like, “Bring your computer.” So we're in Cinque Terre and she hands me a bottle of wine, takes my computer. She looks at my CV, and she highlights all of the talks and all of the publications, and then she hits the ‘delete’ button. I was like, “Ow, that really hurts.” And she was just, like, “Drink the wine.”[laughs] So I had her to help. I also reached out to a couple of female data scientists, and asked them “How do I do this?” (17:47)

CJ: What they said was, “You need to emphasize the skills that you have, and those that you've developed in industry, rather than the things that you've done.” So, instead of talking about, “This is the research I did during my PhD.” I changed it to, “Here are the skills that I learned while doing the research in my PhD.” I didn't plagiarize, but I would read other people's stuff and thought to myself “That also applies to me.” Or like, “Oh, that's also something that I can put on my list.” Then I reformatted it so that the education section was at the bottom – it didn’t matter anymore where I went to school or what my degrees were, rather what the skills that I picked up while I was doing those were. It was genuinely… I had good friends and I was very lucky. A nice bottle of wine helped because it hurt [laughs]. (17:47)

Alexey: I guess most academics have to deal with large datasets at some point. At least in STEM, right? For example, I know that my mother, who works in ecology, needs to process weather data. She does that in Excel. Sometimes she asks me for help to process it. But yeah, I can see how you can reframe this, what she's doing, in such a way that it sounds like what data scientists do. (19:55)

CJ: Yeah, exactly. But I had luck and I had help. (20:35)

Alexey: How many iterations did it take for your CV? You said that you had a CV that was six pages long. You had this bottle of wine and your friend when you were doing this. How many times did you need to redo the whole thing before you started applying to jobs? (20:40)

CJ: The whole thing, I don't know. I did about 14 iterations. As I said, I did this with my friend, Penny. Then I had two different women in industry who just were kind enough to look over my CV. Then I had a few other friends from my cohort during my PhD – my friend, Roxy, who had already transitioned into industry. I don't want to plug her, but she's doing incredible work and she gave me really good feedback. And then I also had my friend Simon, who was also transitioning to industry at the same time as me. As a typical academic, I did the first iteration and then I sent it out to literally everyone and got all of the feedback. I also talked to a couple friends who are recruiters. I was just like, “Hey, can you take a look at this and see what's important?” (20:56)

CJ: Because for a lot of it they were like, “Make sure that you have a PDF version (not .docx). Make sure that the font is something that's machine readable because a lot of people use scanning software to reduce the number of resumes that they have to look at.” So that's why those buzzwords are important. I had LinkedIn Premium at the time. My friend Jessica was like, “You should pay for LinkedIn Premium.” Which is something I almost never do. LinkedIn asked me to pay for premium twice a week and I'm like, “No, I'm good.” But at the time, I actually did, because it allowed me to see the skills that other people were listing as their skills. (20:56)

CJ: I just sat down and was like, “What are my skills? I don't know. I can't do anything.” But when I started looking at what other people were listing as their skills, I was like, “Actually, I can do that. I can do that. And I can do that.” So this allowed me to build up this word skill set, based on what other people had also done, and that I could then attribute to myself. (20:56)

Data science courses

Alexey: Yeah, cool. Thanks for sharing. You mentioned the Coursera specialization. There is a comment in the live chat where people are asking for the name of that specialization. What was it? (22:46)

CJ: Yes. It's old school. It was the John Hopkins Data Science Specialization. (22:56)

Alexey: The one with R, right? (23:02)

CJ: Yeah, exactly. I mean, this was fall of 2017. I think it didn't teach me a lot of the coding things I needed to know, but I think it gave me a good broad overview. Because it's 10 classes, right? It gives you a broad view of the field and all the different things that you can inspect in data. (23:04)

Alexey: Would you recommend this course now? Or would you recommend doing something else? (23:28)

CJ: I think it depends on what you want to go into. I should also mention – I'm a bit of a masochist. I didn't have any money. So I looked at it and the prices were not per class, but per month. Each one was supposed to be one month, so it was like a 10-month class. And I was like, “I can't afford that. I don't have that kind of time and money.” So I looked at it and I set myself a schedule where I did one course a week. That way, I only had to do it for like two months. There was also a four day weekend there, because it was fall break. And so I did it. I managed to get all of them done in like two months so I didn't have to pay for that much. (23:35)

CJ: But I was really focused at the time and really ambitious. I think it was good for where I am now. Because now, I'm leading both data scientists and data analysts and data engineers. And at the time, it was really good, because it gave me this broad view. But if it was somebody today, I think a lot of the field has become a lot more specialized. So I'd ask somebody, like, “What are you interested in?” I think everybody should take Andrew Ng's machine learning course. I think he does just a fantastic job of explaining things. (23:35)

Alexey: This is where I started in machine learning. (24:48)

CJ: Exactly. It's really good. I think he does a really good job of explaining complex things in a very simple way, which, by the way, is one of the most important skills that you need as a data scientist. I think the strongest skill I have from academia is being able to explain complex things in a simple way. That’s really good to emulate. But I know a lot of people are really into deep learning. I think Fast AI has a really great platform for learning deep learning. I think if people are more interested in statistics, there's a couple of Bayesian courses that I think could be really good. So it depends on what people are interested in now. And then I would send them in a different specific direction. (24:57)

Alexey: So you rewrote your CV – you said you had 14 iterations – then you took a course marathon. You finished the specialization instead of 10 months, you did it in 2. Then you decided to move to Berlin, right? You were in Germany, but you weren't in Berlin yet. So you just selected the closest tech hub? Or how did that happen? (25:37)

Moving to Berlin

CJ: It was almost more organic than that. When I moved from the US, it wasn't even on my radar of places I would ever want to live. But living at the university in Halle, which is like an hour and a half south, I would have to go to Berlin to fly out. Halle has an airport, but it doesn't really fly anywhere. So, I spent a lot of time in Berlin and in that year and a half that I was living in Halle, I fell in love with Berlin. It wasn't just that it was the closest city, it was the place that I wanted to live then. I started looking for jobs in Berlin. But I also had, at that point, in my mind a list of places I wanted to be because I really liked the product. At that point, I was already an N26 customer and I was like, “I think this is a great product.” Especially as a foreigner, being able to get a card that works quickly, that you can do fully mobile that… (26:01)

Alexey: That you can speak English to the support and they reply in English? [laughs] Which is not that common. (27:00)

CJ: Exactly. Not common at all in Germany. I liked the product and there were a couple of companies I was looking at. I was talking to GetYourGuide, but I didn't have a reference there, so I was a little bit behind. Then there was a data science consulting company who gave really good talks in Berlin. They had a whole fantastic seminar series. I interviewed with them, but just the initial interview. So it was like, “I love travel. GetYourGuide seems cool.” And GetYourGuide’s blog was amazing, so I was attracted to them. So I only applied to those two companies. (27:05)

Alexey: So you did pretty thorough research before applying. It wasn't ‘spray and pray’. You were pretty selective. (27:40)

CJ: Yeah, exactly. It's so funny, because it's so different when I was leaving N26 – I did the numbers game, the ‘spray and pray’. But when I was applying for my first position, I did all of this research and would read up and look at, “Do these people look happy? Does this founder look good?” I told my friend Jessica this and she was like, “CJ, it's got to be more like Tinder. You just gotta keep swiping. You can't fall in love with the person until they respond. Otherwise, you're gonna get heartbroken by somebody who doesn't even know you exist.” That's the advice that I give now, but that's definitely not what I did. I was just like, “This product is super cool. I want to work with the people who built this.” Then I met the people who were there and I was like, “I really feel like I could learn from these people.” And I was right. I learned a ton. I had amazing colleagues. I got really lucky. (27:49)

Being selective vs ‘spray and pray’

Alexey: Yeah. You keep saying that you were lucky. But if I summarize what you did – you did 14 iterations of your CV. You took this long specialization in two months. You were also very selective – you selected companies that you knew. For example, N26, you already knew the brand. You were a customer and you knew the product. Maybe it wasn't just luck in the end? (28:36)

CJ: It was a lot of luck. But yeah. It was also a lot of work. (29:06)

Alexey: Yeah. At the beginning, you said you just got lucky and that people just helped you. But there is more to it than just that. But now your advice would be to apply to more companies – not selecting a few. Instead, you take a city and apply to all open positions? Or how would you approach it now? (29:11)

CJ: Um, no. I think I would still… I would have increased my numbers. For example, after I was leaving N26 – one of my best friends, who is still the head of data analytics at N26 – he knew I was applying when I was leaving. I got my first rejection. I was like, “Oh, I got rejected.” And he was like, “You're gonna have to get over that. You’re gonna be getting a lot of those.” But at that point, I had been so selective the first round that I wasn't familiar with this feeling. My little sister always says “Find what you love and do it in the evenings and weekends.” And I'm never going to be that person. I was like, “I love what I do. And I'm passionate about it.” (29:33)

CJ: I'm looking for a space where I can be passionate and invested in what I'm doing. Sometimes too invested. Even then, I would go through job ads and be like, “Does this look like a cool product? Does this look like a place I could really pursue and really be invested in?” But in the second iteration, when I applied to like 20 positions, and went through the interview process – I got a lot of rejections. But I applied to many more places. I think the market’s a lot more competitive now than it was four years ago. (29:33)

Moving on to new jobs

Alexey: Yeah. The second job you had was also in Berlin. It was Klarna, right? (31:00)

CJ: It was. Yeah. I moved from one FinTech company to another. (31:04)

Alexey: OK. So was it difficult for you to make the move? (31:09)

CJ: I hate onboarding. But was it difficult to make the move? It was easy to make the decision. The Klarna interview process was a lot longer. I had four interviews back-to-back. I walked out of each of the rooms for the interview and I was like, “I was not the smartest person in that room.” That's a great position to be in when you start somewhere new. If you can start in a position where other people are teaching you things, then your growth potential is huge. So I was like, “This is a company I want to go to.” Because, at the time, there were 36 data scientists and data science and machine learning are really embedded in the product. And so I was like, “This is a place where I can really learn a lot and learn from really cool people.” (31:14)

CJ: But the first two weeks are the same as everywhere, right? Even in my current position, in the first two weeks, you go from a place where you know a lot and you are comfortable in your space because you know the product and you know the company. You know who to ask if you have crop problems. But then you go from that to a space where you know nothing. You have to relearn all of those things and you have to relearn those connections. You don't even know where to go to get the information. The first two weeks at Klarna, it was very much “Ugh. All of my friends are at N26. I knew everything there. I was very comfortable. Why did I put myself in a position where I'm once again uncomfortable?” So, I hate onboarding. But after I got up to speed I really enjoyed my work there. (31:14)

Alexey: And this is how you moved to Sweden, right? Through Klarna? (32:48)

CJ: Yeah. (32:51)

Alexey: Okay. Because they had a different position in Sweden? And this was how you moved? (32:52)

CJ: Ah, no…? There's no way for me to say this without sounding arrogant. I told my manager – she and I are still really good friends – I told her “I really want to try out Stockholm.” I was living in Berlin for a while and I wanted to try Stockholm. She was like, “Okay.” So I started looking for a new position in Stockholm and about two weeks into looking for a new position in Stockholm, she was like, “I can't afford to lose you. I'm just gonna move you to Stockholm. You can work remotely.” And I was like, “Cool.” (32:57)

Alexey: Ah. So it wasn't for a position. There is also an office there. Is there not? (33:29)

CJ: Yeah. The main office for Klarna isn't in Stockholm. So it wasn't for a different position within Klarna. It was for the same position on the same team, but she was willing to move me because I wanted to go. (33:34)

Plan for transitioning to industry

Alexey: Oh, that's nice. Let's say somebody right now lives in a small town in Germany. This person works as a postdoc at some university, doing some science. They hear that data science is cool, so they think, “Okay, maybe I want to do that.” Of course, they want to move to Berlin or some other city. For that person – what should they do? How should they approach from where they are now to a company, or a startup in Berlin? (33:48)

CJ: Yeah. That's a very good question. I think the biggest thing you can do is things like – try to start engaging with the data science community in Berlin. I think it's one of the great things about Berlin – and Stockholm, but very much in Berlin – is that it's such a thriving data science community and people across different companies are super stoked to talk and work with each other. So the first step – or the first piece of advice now would be – go to virtual meetups, or go to them in person if you can manage to be in Berlin. Start meeting people, start talking to people about their work. (34:29)

CJ: In my experience, because it's such a great data science community, most people are super stoked to help. Also everybody's hiring, right? So they’re going to be like, “Hey. I heard you are interested in this. Maybe you should apply to this one.” Or like, “Hey, I'm looking for people like you. Maybe we should talk more.” But I think getting into that community would be the first step, if somebody was working as a postdoc and needed to get in there. A remark I've always said is that I'm remarkably bad at networking, because I have friends who are like, “You gotta work the angles. You gotta meet the people, so they can figure out how they could do the things.” And I’m just like, “I'm not that. I want everyone to be my friend.” So I tend to talk to a lot of people and I have a lot of people in the data science community. But it's never like, “Make sure that everybody's good.” It's just like, “Hi! You want to get a beer? I'm CJ.” I think that can take you a really long way. (34:29)

Alexey: Then, what else? Or that’s enough? (36:07)

CJ: I honestly think that's enough. Because, like I said, I think for entry level data science positions most people are looking for the same thing – they’re looking for smart people who are willing to learn. Once you start interacting with people in the data science community, and it's obvious that you're willing to learn, then I think everybody's hiring – especially right now. I think that would be a good way. Then once you have your foot in the door the rest is history. (36:10)

Requirements for getting hired

Alexey: We also mentioned like half an hour ago when we started, that now, many companies don't just do one interview and hire people. The process is more complicated these days and the requirements are a bit higher. So what do you think the minimum requirements to start the job are nowadays? What do you need to actually know? (36:43)

CJ: Excellent question. I can say from the Klarna hiring pipeline, you probably have to be better at writing code than I was when I started. I think that's the most obvious one. I think the technical test is a lot more arduous, even for really junior people. I think learning and understanding what clean code is, and what good coding practices are for data science – I think that’s going to be crucial. (37:08)

Alexey: And how does one do that? Do you know? (37:35)

CJ: I didn't learn how to do that till after I started. But the way that I learned how to get better at programming – and I still do this – a data scientist friend of mine asked me this today, “How do I get better at writing code?” And I was like, “Find somebody who's better than you are and just pair-program with them every week.” I have a fantastic colleague that I worked with at both N26 and then they followed me to Klarna. They are much better at Python programming than I am. They've been doing it since they were little. So I blocked the time on our schedule for like a solid year when we were at Klarna for an hour every week. (37:39)

CJ: All of the meetings in my calendar are named bizarre things, but our meeting was called ‘stuff and things’. We would meet up and pick one Leetcode problem and then we would pair-program for an hour every week. This way I could learn how he approached things, and I could learn about algorithm things that I didn't know. At that point, my first manager was really strict on clean code and so I learned a lot of clean coding practices from that. Peer reviews, like getting a good peer review process and learning from how other people would do things. I think all of that's really good. (37:39)

CJ: But for me, all that happened after I started in industry. If I was a postdoc, the advice I would give is “Find somebody who's better than you are, who you can learn from, and really go at that.” But that's also something that's really different between academia and industry. One of my colleagues at Klarna was brand new, out of academia, and he did a math PhD. He and I had a long conversation about how he needs to get better at collaborating. Because in academia when you collaborate, it's like, “We're going to do this project together. I'm going to go into this room and do this part of the project by myself. You're going to go into that room and you're gonna do that part of the project by yourself. Then we're gonna smush it together at the end.” And I was like “In industry, when we say ‘collaborate’, it means we're gonna sit next to each other. There's gonna be one keyboard and you're gonna have to feel comfortable looking stupid. You have to feel comfortable admitting that you don't know everything.” I think that's a big barrier to get over. But once you get over that, you learn so much. So that would be my big thing. (37:39)

Alexey: So your suggestion would be to look around and find people who are better at coding than you and pair up with them. Then perhaps offer something in return. (39:48)

CJ: Yep, exactly. Cookies are a great offer. (40:00)

Publications, portfolios and pet projects

Alexey: [laughs] Yeah. And it goes back to the first step, which is networking. Would that be enough? So networking, and learning to code. There is a comment about a publication being in a top data science journal that says “first authors”. Do you think that would be an important thing to do? (40:02)

CJ: Nobody cares. There's one team in Klarna that does research exclusively. They do really awesome research and they're a super fun group of guys. I have an ongoing debate with one of them about the utility of Bayesian statistics. They're super great, but I think most people in industry don't have a background in academia, so they don't care about the publications that you've done. I always look at people's papers if I see that they've published on LinkedIn. Before I interview them, I'll get a background so I understand what they know. But unless you're applying for a research position in industry, that's not a skill that's valued, I don't think. (40:24)

Alexey: What do you think about a project portfolio and pet projects? Do you think it's important to have in order to get a job in data science? (41:03)

CJ: When I'm looking for people, it's not. But I think with other people it might be. I think if the person you're interviewing with has a background in engineering, then I think it makes a difference because they like being able to see what you've done. If I see that, my first thought is like, “I like that you've put in the effort. But that is so far away from real-world data that you're not gonna be able to use a lot of those techniques in real world production.” (41:12)

CJ: I had a junior data scientist I worked with at Klarna and she was brand new. She just started and came out of her Master's. I even had her do the training data. I was like, “Here's a giant training data set.” And she was like, “I had no idea how much of this was going to be cleaning the data. The datasets always come nice and clean when we do them in school.” I was like, “Yeah. No. That’s what it looks like. I'm terribly sorry.” That's why I think I put less value on that. In my experience, a lot of pet projects for data science tend to pull from the really easy to find and really clean datasets, and those aren't always very meaningful. (41:12)

Alexey: The projects you did in academia weren't clean. You had to clean them a lot. That could be a good portfolio project, right? You could say, “Okay, I have this massive amount of genomic data of three or four gigabytes.” Then you needed to do some bash stuff to digest that. (42:24)

CJ: Yeah. Even now, when I have a massive data set and I'm trying to figure out how I need to parse it, I'll usually turn to bash first. But I think that's harder to translate, right? Because if I tell somebody like, “Oh, yeah. You built some de novo transcriptomes on long and short sequencing reads.” That means something to somebody in biology. If I'm going for a biotech job, yeah – it's super important. Put it in there. But I wasn't going for a biotech job, I was going for a fintech job. Those words don't mean anything to the people I'm applying to. (42:48)

CJ: People always tell us “You should tailor your CV to the job you're applying for.” And I tend to tailor my speaking style to who I'm applying for. If I think the person isn't going to understand the jargon, I'm not going to use it. I think that that's worked for me in the past, but it depends on what you're doing and what you're looking at. (42:48)

Adjusting to industry

Alexey: Yeah, thanks. We have a question about when you transitioned from academia to industry, “Did you have to adjust your way of communicating and interacting with colleagues? Was it difficult for you?” (43:44)

CJ: Oh yeah. I really hope that nobody who was at N26 when I first started is in this conversation, because it was incredibly difficult. They will tell you all of the mistakes that I made. But they were mistakes that were easy to learn from. One of them was – I say this to people now, especially if they're giving talks – nobody likes feeling stupid. But it was really hard for me to just… this is so arrogant. It was hard for me to adjust to the idea of knowledge that I thought everybody knew. I’ll use an example where I was talking to a project manager. We were talking about this A/B test he had done and he really wanted to understand whether or not it was significant. So I ran this distilled analysis for him. Then I was talking about p values and he's just like, “Okay, CJ. I need to stop. I have no idea what a p value is.” And I was like, “Oh, okay. Let’s take a step back.” [laughs] So things like that. (44:04)

CJ: But also… everything. Like the whole idea of Slack, I can’t even tell you. I started and I was brand new to the industry, brand new job, brand new city – complete change. I was scared to post emoji responses on Slack to what other people were saying for the first week. I was just like, “Oh my God. What if it's an inappropriate thing? What if I'm doing this wrong?” That was a whole means of communication that I had to learn better. There was so much. But a lot of it was cultural and I think the biggest thing for me is this realization that in academia, everybody has a similar experience. (44:04)

CJ: If you get to be a postdoc, then you know that every single person has done a PhD – there might be some edge cases – but everybody knows what it's like to think, “Okay, I have $2 in my hand and that's all the money I have to feed myself for the next two weeks.” Right? You have this shared experience. And things like, “Okay, I defended my dissertation. It was incredibly difficult. I cried in the hallway.” But everybody has these shared experiences. As you move forward in your career in academia, you have these shared experiences that you can bond over. That doesn't exist in industry, right? (44:04)

CJ: I remember, we were in the bank and we're talking about doing an overdraft experiment, and they're like, “Oh, we'll just issue a couple people 250 euros in overdraft.” And the people in the bank were like, “Who's ever going to use 250 in overdraft. That's no money at all.” And I just looked at them like, “Oh, you've never been poor.” We didn't have these things that were shared, things that you can rely on in terms of the culture in academia. So I had to shift the way I was communicating with colleagues. Not just the tools and not just the idea that I assumed everybody knew a thing, but also the shared experience that no longer existed. (44:04)

Alexey: Somebody is writing in Slack that they're going through the exact same thing with Slack right now. (46:56)

CJ: [laughs] It's terrifying. It was so terrible. All of my colleagues there at the time will tell you that I just walked around with this terrified look on my face for the first two weeks. And I don't have a poker face, so everybody knew. [laughs] (47:03)

Alexey: Also, I think, – in academia, at least – in the places where I was a student or where I visited, it's always rooms. You have rooms where people sit – it's not open spaces. So it's a room with two or three people, but it's always a room. Then when you get to industry, it's usually a floor without walls and people sit together. Was it also difficult for you to get used to? (47:18)

CJ: Surprisingly, no. But I think that's just a ‘me’ thing. I could see how that would be a problem for other people. But when I get in the zone and I'm really focused – the same thing is true when I'm reading a book. There could be bombs exploding around me and I have no idea. When I get focused, it doesn't matter what's happening around me. I also like the social aspect of it a lot. Like I said, one of my best friends is still the head of data analytics at N26. (47:45)

CJ: At the time, the data team was only 10 people. So the data science team and the data analytics team sat right next to each other. Our chairs were right next to each other. Day two, we were immediately best friends. So he made me laugh, and I have a really loud laugh. But he made me laugh so much that, as the data team grew and the data science team ended up in a different space, and we weren't sitting next to each other anymore, like half of the office commented, “Now that Robin and CJ aren't sitting together, everything is a lot quieter around here.” He wasn't making me laugh as much and so the whole area quieted down. So I could see how it would affect other people. But the open space wasn't a weird thing for me. (47:45)

Bad habits from academia

Alexey: Okay, yeah. I'm also curious, from the skills you had before in academia – we talked about many that were actually useful like being able to clean data, to process large amounts of data, and all the statistical things. But what about things that weren't useful? Things that were maybe even harmful? Were there things like that or no? (48:50)

CJ: Yeah. Oh, I was asked this the other and I forget what my answer was. Things that weren't useful. The competitiveness. Like I said, in industry people tend to collaborate more. So hiding what you don't know and trying to compete with your colleagues – I think that’s really not useful in industry. I think that it can be really harmful. But that's something that's innate in academia, because everybody is like cutthroat to beat each other. Yeah. Again, I was really lucky that my team was really collaborative at N26. But even then, it took me a couple of weeks to feel really comfortable looking stupid in front of other people. (49:18)

CJ: In academia, everybody's really smart, but you're also spending a lot of time trying to be smarter. You don't show as many vulnerabilities, so it took me a couple of weeks to really feel comfortable. Being like, “I don't know this. Can you help me figure it out?” I think that potentially can be really harmful. I think the same thing was true of this colleague at Klarna who had just come from a math PhD. It took him about two months to admit that he didn't know something. He really struggled with that and it was the same problem where it's like, “This is not how we were taught to behave.” But this is how you need to behave in order to be successful in this new environment. (49:18)

Alexey: Is that any way to… you just need to do this, right? You just need to be in that environment, learn, and try to adjust, right? (50:56)

CJ: Yeah. The way that I'm doing it now – because I'm facing a similar problem at my current position, where I'm leading people and trying to get them to feel comfortable talking to each other and feel comfortable being stupid in front of each other. The way that it worked at N26 for me – we had this Russian data engineer who still rules with an iron fist, he's fantastic. He read the fine print and found out that you could do a team dinner once a month. There's a budget for it. But once a month, you could take the team out to dinner. The data team was the only team at N26 who did this often. Once a month, we all went out to dinner and just hung out with each other. (51:05)

CJ: I think a lot of it is being able to break down the barriers, establishing that these people are your friends, and that they're not judging you. Then the dumb questions can come and then you can feel comfortable asking that. But I'm doing the same thing right now where there's a lot of knowledge silos at my current company and I’m like, “We're all gonna go hang out and eat dinner. We're gonna go hang out drinking. I'm throwing you guys an event to do this.” Just so that I can get people to feel more comfortable looking stupid. I also talk about how I fail. I try to get people to feel comfortable enough looking dumb in front of each other that they are willing to ask each other questions, even if they think in their mind, “This is a dumb question.” If you don't ask, you won't know the answer. (51:05)

Alexey: Okay, yeah. Thanks. So basically, people just need to be comfortable with each other, get used to each other, and it just takes a bit of time. You don't feel comfortable immediately. You just need to give yourself a bit of time and then things become easier. (52:24)

CJ: I hope so. That's been my experience. And it seems to be slowly working in my current company. (52:41)

Topics with long-term value

Alexey: Yeah, thanks. This is an interesting question from Mateus, “Have you had time to research or explore a topic that may not have an immediate impact on your job, but led to long-term value?” (52:45)

CJ: Like actual data science topics? (53:06)

Alexey: I think so. (53:08)

CJ: Yes. Natural language processing is one of my favorite things and I employed it within other data. But I haven't done many full natural language processing assignments. At N26, and then at Klarna as well, I used that time in the morning to study and learn new things, like trying new Kaggle competitions and learning new techniques. Although I've never deployed a full NLP model, I know that those things are useful. Now I'm leading people, so I'm not doing as much as an individual contributor, but we have a couple of assignments that will be NLP assignments. So I know I can now help other people, my people, to learn the skills that they need so that they can succeed in this assignment. Even though I haven't actually used it myself yet. (53:12)

Alexey: That's quite an interesting perspective. So even though it wasn't immediately useful back then, you still help your colleagues. That's cool. What about Kaggle? Was it helpful for you? (54:07)

CJ: No, because I don't actually compete. I just use it as a source of cool datasets and neat ideas. Then I can play around with those datasets and ideas and try them out. Also, I'll try it out myself first and try to solve it. But I was in the hiring pipeline at Klarna and obviously, I'm hiring now, and it's really interesting to see not just how I answered the question, but how other people do. That gives me a much deeper experience, because then you're like, “I didn't think about it from that angle. But now that I've thought about it from your angle, I've now incorporated that into my own mentality of ways I could approach this problem in the future.” (54:22)

Alexey: So Kaggle was useful, but not as a competitive platform. (55:01)

CJ: Not winning any of them. Yeah. It's really useful to learn how other people are doing things. (55:04)

Alexey: There is an insane amount of knowledge there. There’s just so much stuff there, in each competition. There's forums there and in the forums there are discussions – and there is an insane amount of information there. (55:09)

CJ: Yeah, for sure. (55:27)

CJ’s textbook

Alexey: Um, I know that you wrote a textbook. What was the textbook about? (55:28)

CJ: This is such a funny story. When I saw you put this down in the list of questions. I was like, “Yes, I wrote a textbook.” But it's really funny and it wouldn't have happened if I didn't have another really good collaborator. So I was teaching this course during my PhD. Parasitology was the course – the study of parasites. My PhD is on host parasite coevolution, so it was natural that I would be teaching parasitology. But the teacher, the professor, retired basically halfway through the semester. So we were sort of left in this position where I had to write the course. If you've never written a course from scratch, it's incredibly difficult, and it takes a ton of time. It probably put me back about a semester from graduating. (55:34)

CJ: But, I had this colleague, and I had the knowledge. I was like, “I know a lot about parasites. I've been studying this for the last 10 years. I know a lot about ecology. I know a lot about the evolution of them.” I had some ideas for what we could do for the actual class. So I would write and then she would edit it in such a way that made it friendlier to the students. We put together this whole class on a shoestring budget. We were driving the whole thing. We wrote the tests, set up the website – did the whole thing. (55:34)

CJ: At the end of the semester, because I was a graduate student, I believed that everything I did was worthless. But she has a much better sense of self-worth. So she approached the department, and she was just like, “Hey, we wrote this course. Do you guys want it?” And they're like, “No, we're not buying anything from you. You guys are free labor.” And she was like, “Cool, fuck you.” Then we went to a textbook company – I looked this up the other day – Hayden Macmillan or something like that. I forget the exact textbook company. But she approached them and she's like, “Hey, we wrote this course. We can turn it into a textbook if you guys are interested.” And they're like, “Yeah. Sounds great.” (55:34)

CJ: So next thing I know, Kim and I – Kim was my colleague, she's incredible. Kim and I are on this phone call talking about cover art for our textbook. I was just like, “I don't know how I got here, but this is great.” So yeah – we co-authored this textbook together. Then they set our deadline and I remember that four days before the deadline, we had to take all of the stuff that we'd written and put it into chapters that made sense. We had a schedule. We just sat on her couch – her husband took care of us – we just sat on her couch for four days rotating between pairs of yoga pants every 12 hours while he fed us and did laundry. And we just wrote for four days [laughs]. But yeah, it's “The Evolution and Ecology of Parasitology. (55:34)

Alexey: So what was on the cover? A parasite or something like that? (58:12)

CJ: Oh, it’s this fantastic photo from one of our students of a parasite underneath a microscope. It was pretty cool. (58:16)

Alexey: Slightly unrelated topic, but there is a comment from Aaron, “There are tools for a team to feel comfortable with each other. The tools are liberating structures, serious games that people can practice to acquire trust, learning vulnerability, and so on.” (58:25)

CJ: Totally. (58:43)

Wrapping up

Alexey: Do you want to add anything before we wrap up? (58:45)

CJ: No. I think, you know… I feel very lucky because I've been surrounded… You asked me before, “How do you keep finding good people to give talks?” I was like, “I've been working with awesome people that are willing to teach me cool things.” In that regard, it's taken a lot of hard work and I work long hours often. I'm super stoked to be teaching people now. But I also feel very fortunate that I've encountered so many awesome people and that I get to do something that I love. (58:48)

Alexey: Yeah. Thanks for supplying speakers for DataTalks.club. (59:19)

CJ: Yeah, anytime. (59:23)

Alexey: So if you have more speakers, please let me know. [laughs] (59:24)

CJ: I always have more speakers. (59:26)

Alexey: Okay. Thanks. So, how can people find you? (59:30)

CJ: I'm easy to find on LinkedIn, CJ Jenkins. I think I'm almost always the top choice. But yes, CJ Jenkins on LinkedIn. (59:33)

Alexey: OK. Thanks a lot. Thanks for joining us today. Thanks for sharing your experience with us. And thanks, everyone, for being active and for asking questions. There are quite a few comments in the live chat. So thanks for being active and have a great weekend. (59:42)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.

DataTalks.Club. Hosted on GitHub Pages. We use cookies.