What are the different paths toward a career in data science? What do you wish you had known earlier? How do you “own” your career? Data scientists from Uber, Facebook, and Reddit answered these and many other fascinating questions during a discussion we hosted last week with Uber’s Women in Statistics, Data, Optimization and Machine Learning (WiSDOM) employee resource group.
The full recording is below, but this blog has some highlights from the event, Women in Data Panel & Mixer: Build and Advance Your Career in Data Science.
Sreeta Gorripaty, senior data scientist at Uber: So the first question that I love asking when I meet really exciting people is: what’s your story? How did you get here? What was the path that took you here? Were you like, at the age of 20, 21, 22, “I want to be a data scientist. I’m going to be that.” How did you get here? Aude, let’s start with you.
Aude Hofleitner, data science manager at Facebook: Sure. So, for me, it wasn’t like I wanted to be a data scientist when I was 21, 22, mostly because when I was 21, 22, data science wasn’t really a thing. I feel like for me there’s been a lot of coincidences that have brought me here. I mean, I studied in France—that’s where I’m from—where I was doing mostly applied math and engineering studies, but mostly on the theoretical side. And I was also really interested in transportation. And so I went to Berkeley for what was, at the beginning, an internship in a transportation group and then turned into a Ph.D. in machine learning, but still applied to transportation.
During my Ph.D., one of my Ph.D. advisors was like, “Oh, you should try an internship, pick up some hacking skills and things like that.” And so I interned at Facebook. I don’t know if I picked up much hacking skills, but I realized that there were a lot of things that were really interesting for me to do there. My first intuition was like, “Well, I’ve been doing mostly applied math and what else would I do besides academia?” And then I realized that there was actually a lot of things that were a really good fit for me and for the intellectual stimulation I was interested in.
So I came back from my internship and I was like, “Well, actually I don’t think I’m going to do academia.” And there’s a lot of really exciting things to do for me in the data science team at Facebook and that’s how I landed there.
So, it’s something that I’ve really enjoyed as it’s combining the research component that I came from as well as solving real-world problems, like having an impact that’s measured on the world rather than the number of people who read my papers.
Marie-Camille (MC) Achard, data scientist at Uber: I’m also from France. I think I had the same curriculum at the beginning, so I liked mathematics in high school. I went for applied mathematics and physics and then I got into the—we have engineering school there. The system is a bit different than the U.S. and we get to try basically every type of engineering, so you do a little bit of mechanical engineering, civil engineering, financial engineering, etc. I still liked mathematics, so I went for financial engineering. Then I decided that it was too theoretical. So I tried six months of investment banking, which wasn’t theoretical enough. So I went back to the mathematics and data fields. That’s how I arrived in San Francisco. And I loved it so much, I stayed for a master’s degree at UC Berkeley, where I got to develop more of my machine learning skills, my coding skills in general. And I could see through the projects how you can get a job where you get to do theoretical stuff in applied real-world problems, which I loved, and I always liked transportation.
And I think at the time I thought Uber was the perfect place for me. I guess I ended up at the right place. Yeah, it’s been three years and I still love it.
Katie Bauer, senior data scientist at Reddit: I didn’t major in any mathematical field. I was a linguistics major, and I particularly was focusing on a subfield called quantitative sociolinguistics, which is like using social variation to model the way people talk. And I thought all of the quantitative stuff in that was super cool. I loved getting my feet wet with R and I wanted to find more ways to do statistical modeling while still staying within my linguistics field, which led me to a subfield called computational linguistics, which a lot of people now know as NLP.
But I was finishing school and I didn’t really know what I wanted to do. I didn’t really want to spend more time in academia. So I thought I would move to the Bay Area and work as a computer programmer because I liked programming and had done a bit of that in my coursework. I could not convince anyone to hire me as a developer, but I did, in a weird twist of fate, end up getting hired as a linguist for a natural language search startup. And I worked a ton on their query parser and in the course of doing that job, picked up a ton of other data-related roles. I did analytics for them, I would build scripts to automate their Twitter feed, annotate data sets for machine learning, and started doing a little bit of machine learning stuff by the end as well.
That company ended up going out of business and I just decided to pivot into a more data-related role generally. So I spent some time in ad-tech as an analyst and then later as a data scientist. After a couple of years of doing that, a recruiter from Reddit reached out. And I’ve been there ever since.
I get to do a bunch of really cool and different stuff at Reddit. We got lots of weird data that’s weird in that it’s confusing but also delightful. There’s just no shortage of interesting problems to get to work on. So I really like being there.
Sreeta: Katie, the next question’s for you. You’ve been a founding member of data science teams and a lot of best practices at Reddit, and you’ve obviously heard this a ton of times, that data science is the most promising—there are other objectives for that—job of this decade. And some people say even the century. When people come to you and say, “Hey, how do I get into that? How do I get that promise going?” What’s your answer for what’s the most optimal path? And I think the bigger question is: is there really an optimal path to be a data scientist?
Katie: Yeah, I don’t think there’s one way to be a data scientist, first of all. I was just saying earlier that I’ve been having conversations with people about, “Are you a data scientist or are you a machine learning engineer?” And we’ll be talking about it and then realize halfway through the conversation that my definition of a data scientist is their definition of a machine learning engineer and vice versa. No one can really agree what these fields mean. And I really think that one day the title “data scientist” will be something like “webmaster” where it’s like, it was super generalist and it just ended up specializing to a bunch of different things.
So with that background, how do you become a data scientist, this very vague thing? I think it depends on what you want to do. There are a lot of ways, a lot of paths up the mountain. There are great bootcamps now. There’s honestly a lot of just degree programs popping up and I’ve had really good experiences working with interns from… like, University of San Francisco has a great master’s program. Galvanize no longer exists, but they had a fabulous master’s program.
The path is becoming way more paved and it’s being more established as a field. So there’s going to be more standard ways in. But if you are already at a company and you’re doing something that’s data science adjacent, I really encourage you to explore that as an option within your company because domain expertise is a very important part of being a good data scientist. And if you’re already familiar with the product that you’re working on, as a non-data person, that’s going to be so much easier to transition. You can always teach skills. Domain expertise is one of the hardest things about ramping up a data person.
Sreeta: Continuing on that, you mentioned this transition in the domain from a maybe slightly less data role to a more data role… how daunting is that? Have you seen people do that? How does that work?
Katie: Yeah, I mean I did it. I was an analyst before I was a data scientist. And it was scary. It always felt like people knew more than me. And I still feel like that all the time. It’s one of those things where you have to spend time realizing that you imagine what you know is a subset of what everyone else knows. But it’s really like a Venn diagram where you have some overlapping area that you both share, but then you both have separate areas, and everyone brings something to the table.
It’s also something where it’s like, it’s super important to be humble about what you don’t know. I work on stuff all the time now where I’m a data scientist and people respect what I say but it’s like, “I don’t know anything about this part of the product. I really need to talk to someone who understands it before I start making assumptions about what this data means.”
Sreeta: So that actually, the not knowing, brings me to the next question, which is: data science today is such a dynamic field. There’s just so much changing: the programs, the algorithms to be used. What works, what’s optimal, is constantly changing. And in such a dynamic field, if you want to keep growing as a leader, how do you stay on top of these trends?
Aude: Yeah, I mean it’s really fascinating and I think if you have a curious mind, it’s definitely a great field to be in. For me, there are maybe two main ways. The first one is that I stay pretty close to the academic field by going to conferences, reviewing papers, and so on. And the other thing is we’re surrounded by amazing people. All my peers and the people on my team all bring different things to the table. I really agree with the Venn diagram image of we always feel like, “Oh, these people know so much more than me.” But you also know so much more than them, just in different ways. Learning from peers is the best ways to learn, at least for me.
MC: I think it’s scary because you arrive with a fixed set of skills and you want to expand. So I guess that conferences are also a very good way for me to learn. Recently, for example, I was at ICML. Do not be scared if you go to conferences. You won’t get the whole thing. There are so many tracks, so many talks. I don’t think anyone understands every talk, just pick a track and dive deeper into those new things, those new papers. I think that’s a really good way of blocking out several days just for learning.
You can also block a couple of hours a week to read papers, read articles. If you’re more into theoretical stuff, then go to papers. But also I know many companies write articles on what they just developed. I know Uber does, but other companies like DoorDash does, Airbnb does, Lyft does. There are many companies that just describe what they do, how they did it, and what were the mistakes, the good points they learned. That’s really helpful, especially to broaden the set of skills.
I think my answers are actually pretty similar, but do not hesitate to reach out to other people that are experts in the new domain you want to dive into. There will be people that will know and they usually are very happy and excited to share what they know.
Katie: Yeah, I will definitely second the—if someone is an expert in something that you want to learn about, reach out because usually no one wants to talk to them about that thing and they’re so excited to have someone to talk to. But in addition to conferences, which are great, there are lots of really cheap ways to learn about data science, tons of good podcasts, a bunch of really good blogs. I love email newsletters, personally.
Think about what resources work for you. Podcasts are great for me because I can listen to them when I’m going for a run or when I’m commuting and I learn a lot. Twitter is great. There are tons of people talking about data science and statistics on Twitter and it can be super quick. You can just check in, see what someone’s thinking about something. And that fits very well into my life, personally. I don’t always have time to sit down and read a paper in depth—I like doing that when I can, but there’s also shorter-form ways that I can get little bites and then that plants a seed in my head and then later I’ll know where to go if I need to find that information.
More broadly, I would also advise that you can’t know everything, so don’t feel bad if you don’t. Two techniques that I rely on a lot in terms of figuring out what I need to learn next are: what do I need to know for my job and what am I interested in personally? Right now I’m learning about Flink streaming technology because I need to use it for something I’m working on. It’s probably not something I would be reading about on my own, but super relevant to what I’m doing. And then in terms of just something I’m personally interested in, I saw someone on my team using a zero-inflated Poisson model and I was like, “I don’t really know what that is.” So I found a YouTube video on it. It was from some longer course on statistics that some professor had posted their entire course on YouTube. And the course was great. And I have spent the past couple of weekends just watching videos from that course. And it’s been super interesting. It’s not necessarily relevant to my job, but I’m learning a ton doing it.
Just having joy about the things that you’re interested in and pursuing them is honestly a great way to make learning sustainable.
Sreeta: Yeah, I totally agree. I feel like one big difference in today’s world is there’s just so much open resources. If you want to learn, if there’s that intention, there’s just no stopping you in terms of how you can do it. I love that.
One of the trendy phrases that I keep hearing all the time, it’s become super common lately, is: own your career. So Aude, you have an impressive background. You’re managing a big team now. What is “own your career” for you? What is the advice you want to give women to own their career?
Aude: I think the biggest thing is: you know yourself the best and you should be active in figuring out where you want to go and how you want to get there. And in some sense, you’re not there on your own and there’s a ton of people around you, but in some sense, let them help you and give them hints about the things that you’re interested about. And then, that’s the best way for them to know that when there’s an opportunity that comes up, it’s like, “Oh, she said she was interested in getting more experienced with public speaking. Maybe I should invite her to that event.” Or, “She said she was interested in getting more into leadership or into this really deep technical field that she wants to learn about.”
The more people around you have heard about things that were exciting to you, the more they’re gonna help you build your career.
Sreeta: MC, I know I’ve personally worked with you multiple times on this. We’ve gone to recruiting events together and I know you’re super passionate about data science culture in the company at Uber and the community. So, what do you wish that we, the data science industry, did differently, and what are some of the practices that you want to encourage some of the future leaders we have in this room to think about more proactively as they take on their roles and grow?
MC: One thing is alignment. Don’t get siloed into a subteam on something. Share it with the other teams, discuss it, get everybody aligned on your project. I know it’s not easy at Uber, we’ve been sometimes failing at that, every team would define something a different way or build an algorithm that would do in the end the same thing but another way. And it takes a lot of time afterwards to go backwards, get the definition that everybody can align on, and build that. So alignment is very important.
And something related: communication. So, developing your soft skills as well. It’s again very easy to get siloed and develop something very technical. But it’s also hard in a different way to share it in layman’s terms with a different audience. So the whole communication side of things is not to be neglected, especially in the industry because that’s also what’s going to get people excited about your projects. And yeah, it’s not necessarily something you learned in school. So, soft skills are also very important. And networking with people, discussing, sharing. Training yourself to present, that’s also very helpful actually.
Sreeta: Time for some cheesiness. I love asking this question because I think it’s pretty insightful. So, if you were to go back in time and look at your young self, getting into this field and getting through your paths, what would be the one advice you would give yourself? Or maybe another way of thinking is, what was the most helpful tip someone gave you?
Aude: I’m sure there’s a lot of things that I could have benefited from as early advice. I think one that I still can benefit from is, “What would you do if you aren’t afraid?” That’s one of the sayings that we have at Facebook. We have a bunch of inspirational posters and this is one of them that Sheryl Sandberg also really likes. I think for me it’s really like, “What are you preventing yourself from doing just because you put these barriers around yourself?”
Imposter syndrome is one reason. There’s a lot of reasons behind that. I feel like it can come in lots of little ways that eventually may slow you down. For example, “Oh, should I reach out to that senior engineer, he looks pretty busy, he’s not going to have time for me.” Or, “Oh, can I email this person who has director in their title? That sounds really scary.” Eventually, they’re all just people like us. They all want the success of the projects. There are titles and there are managers and hierarchies just for organizational reasons, but it’s not that at some point you become a manager or at some point you become a director or VP or whatever and you become a different person. All these people, the reason they got into these roles, is usually because they love people and because they want the success. And so I think for me it’s just like, yeah, what would I do if I wasn’t afraid? And reaching out to people is just one small example.
MC: I think imposter syndrome was also very—I think most of us experience it at some point. At the very beginning, I was almost apologizing for my age and my very junior role. I’m like, “OK, please know that I’m new here.” And then people told me, “But no, you aren’t new anymore. And you gave very valid answers before, we want your opinion.” So you have to own it, really, and from the very beginning tell yourself that you own this, that you are here for a reason and you can do it.
So that would be one of my answers. The other would be to keep in touch with people that you meet along either your studies or afterwards. Just keep in touch with them. That’s what’s going to broaden your skills again and expand what you know
Katie: This is maybe a little redundant with everyone else’s answers, but I agree with them so much. I would tell myself, “Work hard, be nice, and be bold.”
Work hard because that’s what you’re there for. Hopefully, you’re doing it because you love it and you’re getting something out of it. You enjoy it. You find it interesting. Take joy in what you do.
Be nice because the people you work with, they’re going to do work for you. They’re going to be the source of your next job. They’re going to be people who help you get a promotion. They’re going to be people who—they’ll get you on the project that you want to be on. And they’ll just make your day pleasant. Some of the best parts of my career have just been having a fun lunch with one of my coworkers.
And then finally, be bold. I feel like the biggest mistakes in my career have been things that I didn’t do. You’ve got to tell people what you want because then they can help you get it. Never feel bad about wanting something because you probably deserve it.
This transcript had been lightly edited for clarity.
Inspired to start or grow your data science career? Check out our Data Science Career Track—you’ll learn the skills and get the personalized guidance you need to land the job you want. Are you curious to read about amazing work done by women in our Data Science Career Track? Here are two whitepapers on How a Machine Learning Algorithm Helped Make Hurricane Damage Assessments and Revolutionizing Photography with an AI-Based Image Classifier.