At Course Report, we talk to tons of data science bootcamp graduates – most bootcampers are going to a data science bootcamp with the goal of getting their first job as a data scientist when they graduate. And that's most likely going to be a junior level job – a Data Analyst or a Junior Data Scientist or Apprentice. That first job is definitely one milestone in a very successful data science career, but how do you become a Senior Data Scientist?
Today I am joined by two graduates of Metis, which is a data science bootcamp in Chicago, New York, San Francisco, and Seattle. Emily and Halle both graduated from the same Metis class and have now been in the real world working as data scientists since 2017 in San Francisco. They’re going to share how they’ve grown their careers beyond the first job.
Emily and Halle, thank you so much for joining us. Could you tell us what you were up to before you attended Metis?
Emily: I'm Emily and I'm a data scientist at DrivenData. Before Metis, I was working as a research assistant at Stanford. I had done my Masters in international development and I'd been working in policy research for a few years, first at The Brookings Institution out in DC, and then at the Stanford Center for International Development. My work was largely in the economic development space.
I'd been working with data, but the economist’s tool of choice for that is Stata, not Python. I was familiar with data and running some regressions and thinking about communicating results, but I didn't have any of the CS background (so I taught myself Python a few months before applying to Metis). Basically, I had been on a research track but decided that I didn't want to do a PhD. I really loved working with data, but felt like I was pretty limited in the sort of data that I could work with, and so I wanted to go to Metis to essentially fill in that gap.
Halle, what were you up to before Metis?
Halle: Before I went to Metis, I have been working in environmental consulting for three and a half years. I got my masters in environmental engineering. So I was doing engineering consulting for a company that did pretty analytics heavy work. I was doing a lot of data analysis and I was really happy doing that. But I sort of had this consistent feeling that there were more insights that we could extract from the data that we were working with. We didn't really have the skills, and honestly, the whole industry didn't really have the needed skills. It wasn't really a thing anyone else was doing. So I decided to be the person who learns how to do that.
Did you research or consider any other data science bootcamps? How did you end up choosing Metis?
Halle: I did do some research on other programs. The reason that I ended up picking Metis was threefold.
I was looking for something with the correct vibe. I came from academic institutions that were very driven and very intense. I obviously had no issue with intensity but I really personally believe that the way that you learn is not by killing yourself with the amount of volume of stuff that you're trying to cram in. You need rest, and you need food. I wanted an institution that understood that you learn more when you have a balance in your life. So I wasn't looking for an institution that was trying to sell me on how hardcore their program was, because I didn't really feel like I had anything to prove in that respect. That was one point in favor of Metis.
Secondly (and thirdly) I really felt like the team at Metis was interested in their students as individuals, not as numbers to turn through some sort of success/outcomes report. At the time when I wanted to go back into the environmental industry, I really felt like Metis was interested in supporting me and my career goals rather than just trying to shoehorn me into whatever job would make their thing look the best.
Emily, did anything else stand out about Metis in your research process?
Emily: I read some interviews with Jason Moss and I really liked that when he talked about things that really matter in terms of being a good data scientist, he talked about things like perseverance, and curiosity. To me, it feels like it matters more how you deal with what you don't know than what you know, so this focus on characteristics resonated with me.
Halle, you mentioned looking for a “vibe” – did you go to an info session or an event in the classroom?
Halle: I did, and actually I was lucky enough to get one of the info sessions that Jason was presenting himself. So I did get a similar perspective as Emily got about Metis' values as a program. I think it was really beneficial.
Rewind a bit – what did you think was going to be your outcome after graduating?
Emily: I thought that I'd go back to working at Stanford. I'd actually taken a leave of absence from my job because my expectation was that I would come back into a hybrid research/data science role. Part of my role at Stanford was writing pieces for the website that communicated findings from research, so I thought I could pull in data visualization to help with that communication or potentially work with some more interesting datasets if they were available.
But once I delved into data science at Metis, I found that I just loved what I was doing. And I thought, "Man, I just really want to do data science full time." I think about halfway through Metis it hit me that I was actually learning enough where I felt I didn’t have to go back into a hybrid role. If I wanted to do data science full time, I could apply for those jobs. So that's what I did.
Halle: I envisioned myself going back into environmental consulting as an industry and that's actually what I did do. My first role out of Metis was doing data science for an environmental consulting firm. I definitely sort of had this vision of bringing those skills that I learned in Metis back to the industry I came from.
The Metis curriculum covers Python and machine learning, databases, SQL. When you graduated from Metis, what did you feel most confident in and what did you feel like you learned at Metis that you didn't know before?
Emily: Yeah, absolutely. Everything I knew about machine learning when I graduated I learned at Metis. I didn't have that CS background before. I mean, even just the code that I could write at the end of Metis was drastically different from going in.
The thing that I felt most confident doing at the end of Metis was that day-to-day work of data science – of cleaning and munging data, doing exploratory plots, doing feature engineering, training a model. That’s actually still the bread and butter of the work that I do now so I was able to start using those things pretty soon into the job.
Halle: One of the skills that I felt most confident in coming out of Metis was picking the correct algorithms for the problem at hand.
One of the things that I really liked about Metis is that we did end-to-end projects the whole time. We were able to pick up projects and say, "Okay, here's a bunch of dirty data, let's get it cleaned and into a model and so on [do data exploration, validating and scoring the model, iterating on better versions of the model, and then presenting the model]." I was grateful that I didn’t just know piecemeal parts of the process.
I think that is important – I get a lot of questions from people wondering if they can learn data science on their own; but it sounds like the bootcamp structure at Metis was helpful to you.
Emily: I think there's also something different, at least for me, about having that time pressure on me. We had two weeks for a project and had to meet that expectation. That would have been tough for me to enforce on my own if I was just doing this in my free time.
It just felt easier to be able to rely on the team at Metis who had already done so much curriculum development work, distilling the concepts to then teach them.
Halle, you mentioned that you went back into environmental field after you graduated from Metis. How did you get your first job after Metis?
Halle: I had actually been in touch with this company before I applied to Metis. When I was planning on moving to San Francisco, I was looking for a standard environmental engineer position. After I graduated from Metis, I reached back out to that company and asked if they were looking for data scientists. They got back to me in under 24 hours and said "Yes, please come in for an interview." That kind of fell into my lap, which was nice.
At my first job, I was handling a lot of the pipeline ETL for ingesting their data and getting it into a usable form. I did a fair bit of database engineering rather than data science – a lot of visualizations, a lot of data presentation.
Emily: My first job was at the Gates Foundation. I was a data scientist on the agriculture team.I was actually the first data scientist on that team and I don’t think I appreciated or understood how unique of a position that was going in. The work I did was a lot of data preparation work – trying to get the data in one place and making it play nicely together to then build some simple visualizations, less so actually getting to some of those higher levels of analysis. They'd hired me as a data scientist but since they didn't have business intelligence analysts in place, I ended up doing a fair amount of that role.
Halle: That was something that I forgot to add also. I was also the first data scientist at my company. Not only in the office of my company but in the entire multinational, multi-thousand people version of my company.
Looking back, do either of you recommend taking a job as the first data scientist at a company?
Halle: I think it depends on the company. First of all, if you like database engineering and ETL work, then you will be a lifeline at the company and everyone will love you forever. Honestly, in order to do data science, you need data. And if you don't already have the data and the infrastructure to sort of have the data in a usable form, then that needs to happen first before you can ever do anything with the data.
And so, you could maybe be the first data scientist at a very small startup and end up doing a much wider range of work. But if you're the first data scientist at a large company that already has a bunch of data that needs to be pulled into a usable format, then you're going to end up doing a lot of pipeline work. That wasn't really exactly what I was looking for. So that’s a question that I would ask myself.
If you’re interviewing to be the first data scientist at a company, ask your interviewer, “Do you have any database engineers?” and/or “What's the status of your ETL and pipeline?”
Emily, what were those job interviews like for you right after you graduated from Metis? How did they go?
Emily: Interviews were different by company. I interviewed with a startup that was working in education technology, and I was interviewed by their chief data scientist. That interview was pretty standard – phone screening, take-home assignment, you discuss that assignment in interviews, you talk to everybody at the company.
My interview for the Gates Foundation was pretty different. Because I was the first technical hire, the interviews were a lot less technical. I didn't actually do a coding challenge for that interview.
They cared a lot about subject matter expertise as well. So the fact that I could speak the language of agricultural development and then also be able to translate some of those concerns into data was really relevant. Looking back, that non-technical interview process was, in some ways, a red flag. As Halle was saying, the question to ask when you're interviewing for first data scientist positions is, “Who are the other data team members and what are they currently doing?” You want to understand if you’re filling in for a position that they don't have yet or if you’ll get to build on something that they've already put in place.
Are either of you still with the company that you started with after Metis?
Halle: I am not.
So how has your career grown since you graduated over the past 1-2 years?
Halle: I spent about a year at my first job and then I decided that I would really like to move to a company where I could work with other data scientists. I looked for a company whose data science program was a bit more mature and had a set of best practices for doing data science in an effective way.
That was actually some advice that I got from an industry mentor – she recommended that if you want to learn how to do data science well on your own, you need to know what problems are normal and what problems are not. Some problems are inherent to a job and other problems indicate that something's wrong. And to know the difference between those problems, you have to work somewhere that's functional and knows how to do data science. So that what I did. I think it was great advice and so I'm passing it along.
I currently work for Lyft on their Data Science team – that’s the role I got a couple months ago.
Amazing. Emily, what are you up to now?
Emily: Last April, I switched jobs and I'm now a data scientist at a small company called DrivenData. I continue to work in this intersection of international development and data science, which has been a theme throughout my career. At DrivenData, we work exclusively in the “for good” space, meaning that we work with nonprofits and international institutions like the World Bank, that have interesting datasets but may not yet have that in-house, technical expertise to bring the power of data science to the big societal problems that they're working on.
Halle, you mentioned that you had a mentor that guided you in your career. Emily, did you have someone similar?
Emily: I had a great chat with a Metis instructor, actually. I was doing a Q&A webinar about Metis and afterwards, we got to talking. She had been the first data scientist at cars.com and had built out that team. It was really helpful to talk to somebody else who had been a “first data scientist” and had figured out what problems are normal vs. what problems are not normal.
Hearing about her experience helped me frame some of my thinking around my decision to stay at or leave my first job.
Aside from resume changes, how do you both feel like you've grown and changed as data scientists over the last two or three years? Do you think you've become "better" data scientists as the years have gone by?
Emily: For one, my code has just gotten so much better, cleaner and more efficient. There are all these style conventions that I didn't know until I joined my current team. It sounds silly, right? But my boss would painstakingly point out the spaces after equal signs that weren't supposed to be there during code reviews, and it really helped me internalize these norms for writing good code. Now if I look at my code, I just think, "Wow, this is so different." When you're going through Metis, you're doing the best you can, writing code in whatever way works. It's cool to now have the knowledge to be able to say, "Oh, I know that this is going to be a better way to structure this or I have this complicated thing and I think I know how to write the function to achieve that."
Halle: Certainly, there was a point when I was preparing for my second round of interviews after leaving Metis where I went back and did a bunch of the early challenges that we got at Metis. There was one that I remember just bashing my head against forever and ever and ever. I went back and did it again, and worked it out in half an hour. I could tell that this wasn’t the way it went the first time. I think that that was a really good marker of how far I've come.
To be honest, personally, I feel like the bigger change has been in my own confidence. One of the things I really love about my current role at Lyft is that I work for a data-driven company, so the product and engineering teams look to the data team to help make their decisions. I think that's really given me a voice and a confidence to use that voice in a way that I didn't really have before.
What are you both working on day-to-day as data scientists now?
Halle: The one-sentence summary is that I work with the product and engineering teams at Lyft and help them make really good decisions and guide their roadmaps in a data-first kind of way.
Emily: Working at a small company and also a consultancy, it means that we get to work across a whole bunch of different subject matter areas. My work, thematically, is really different. I've been on projects for financial health, energy efficiency work, conservation work. For example, our project on conservation was focused on identifying animals in camera trap videos. In each of those projects, I’m following a workflow that you'd expect from a data scientist: we have a dataset and there's some sort of modeling that needs to happen as well as all the pieces that come before and after that.
In terms of growing as a data scientist, I continue to see different types of data and run into new problems where I end up researching a new method or figuring out how to do new things.
As your data science careers mature, do you find that you’re tending to specialize in something specific? Or are you staying general?
Halle: One thing that can be really valuable is subject matter expertise. Any subject matter expertise you have is going to really give you an edge because if you're developing solutions for an area that you have zero subject matter expertise in, you're not going to create any solutions that are worth anything. You're just going to create nonsense.
I've seen that happen before – you bring someone onto your team who is a fantastic data scientist with absolutely no knowledge of the subject matter and they end up creating something that no one can use. Specializing to that extent is helpful.
There are also just things that you get interested in and start to work more on. But I haven’t really focused on one thing in the first couple of years as a data scientist. I think that's more of a 5-10 year career goal.
Emily: Similarly, I don't feel like I've “specialized” but there are certain things that I tend to find myself doing less of. For example, I've never worked with actual big data where I’d need to use something like Hadoop, and so I don't really have the skill set for that. I also don't tend to do a lot of Natural Language Processing (NLP) because there tends to be fewer datasets for that in the space that I'm in.
If you compare your first job interviews after graduating from Metis to your second round of job interviews when you're getting your second job, how did those change? Do you feel like you got more confident as you went on?
Halle: My first round of job interviews was obviously very different from my second round because I was switching industries in the second round and wasn't in the first round. I'm not sure that that's terribly helpful or applicable to anybody else. I would say in terms of what you're preparing for with interviews, prepare for something that might be a little demoralizing. You're going to interview with a lot of companies and most of the time, you're not even going to make it through the phone screen. And that's okay and that's normal.
There are a lot of companies out there and sometimes it's just a numbers game. Sometimes you're just going to get someone grumpy and having a bad day and that's not reflecting necessarily on your skills. You should obviously take a look at your skills and make sure that you can answer the questions okay and stuff like that. Sometimes you just get asked a weird question and I think it’s important to not take that so very personally.
The advice I got from somewhere online was that you should imagine up front that you have a preordained number of days it's going to take to get your job. You don't know what it's going to be. It could be 20, it could be 200, but there is a number at which you're going to get a job on the other end. Your role is just to make that number go down by one every single day. Every day that you spend job hunting, every day you spend preparing is going to make that number go down by one and you're just going to roll with it.
Emily: It's interesting because I feel like I had a lot of imposter syndrome interviewing for my second job because I didn't feel like I was like fully utilizing all the stuff that I learned at Metis in my first job. And so, when I graduated from Metis, I was like, "I just built a neural net and I've been doing all this fancy modeling," and then over the next six months I really didn't do much machine learning. I was not really doing anything with models. And so when I went to interview again, I had this fear of, "Did I forget this skill set?” I remember Andrew (a Metis Sr. Career Advisor in San Francisco) saying, "It hasn't been that long. It's all still there." And it was.
But I think the thing that stuck out to me in the second job interview was just that I felt like people took me a little bit more seriously because I had already been hired once as a data scientist. We'd been told that getting your second data scientist job is easier than your first and I did find that to be somewhat true, whether it should or shouldn't be. But I think that there is something about the fact that somebody has vouched for you and hired you. Having that on your resume is just a good foot in the door.
If I were to frame that as advice, I think it's more just that it's okay to take that gamble on the first job. And if it's not quite right, you're still going to learn a lot about the things that you want in your next job. And so it helped me really target the companies that I wanted to work for and form the questions that I wanted to ask because I knew the things that I valued, especially if they were things I didn't get the first time around. So it's okay to take a risk and to switch jobs to get to something that you love.
At what point over the last two years did you truly feel like data scientists?
Emily: Maybe when I bought a 32-inch monitor for work?! Not really – great monitor, though. I think it was just noticing that the programs that were always open on my computer were Jupyter notebooks, a terminal window, Sublime and all those different things that you'd expect. That just wasn't the case in my first job. Working in a job where I go to work and I write code all day – that's what made me feel like a data scientist.
Halle: I have two completely diametrically opposed answers to this question. I think this is the type of question that deeply, deeply depends on where you're coming from. And people come from such vastly different angles. I think some people have a failure mode where they never feel like a data scientist, and some people have a failure mode where they feel like a data scientists way too early. I don't want to give the wrong answer here so I'm going to give both answers.
If you're sitting there thinking, "I will never be a data scientist. I'm not one now, I won't be one in the future," the answer is: you're already a data scientist. That was the answer that Andrew, our career advisor, gave me during the third week at Metis. We went to a networking event that said it was "for data scientists" and I was sitting there like, "But I'm not a data scientist. How can I go to this networking event?" And he said, "You're already a data scientist. You're going to this networking event." And I went to the networking event and everyone at the networking event was like, "Yeah, you're a data scientist." So it might have been then.
The opposite answer is to take a couple of years looking at the work that other data scientists are doing. You're standing on the shoulders of giants – figure out who those giants are and the work that they're doing and spend some time paying attention to what's already out there, or you end up reinventing the wheel a lot.
Is there anything that you have actively done to grow from junior to mid/senior-level, data scientists?
Emily: One of the things that’s helped me grow the most is having a team that does pair programming regularly. Yes, it may feel embarrassing at first because you think, "I'm going to forget how to do things and they're going to see me make mistakes." But you learn so much. That would be a question that I would ask in interviews: "Does your team do pair programming and how often and how does that happen?" It's just a really great way to learn from other people's code and knowledge.
Halle: Emily that's a really good idea and I'm going to go ask some of my coworkers if we can set this up!
Pursuing personal projects, especially if you feel like you’re not able to utilize the full range of your skills in your job. I spent a pretty grueling couple of months doing a 40 hour a week job and then 20 hours a week of personal projects on the side, preparing for my second round of interviews. I think that was hugely instrumental in not only having a great portfolio for people to look at, but also developing my skill sets for the interviews and just building general confidence.
Also if you feel like you're getting questions in interviews a lot that you don't really have a great background in, it's usually a good idea to brush up on those skills independently.
Before we wrap up, any advice for folks that are just about to start at a data science bootcamp like Metis?
Halle: I have actually referred a friend or two to Metis. Metis has a bunch of prep work, so my first suggestion is to do their prep work and everything will be fine. I recognize that's not helpful if you're not going to Metis in particular. Regardless, ask your bootcamp if they have some prep work that you can do. I'm sure they'd be absolutely thrilled to give you those resources.
Before you start, brush up on Python. A Python 101 course would be really helpful, and a lot of math and statistics. If you know your statistics backwards and forwards, you're going to save yourself a lot of time later trying to wrap your head around various ways that certain algorithms are behaving.
Very practical advice. Emily, anything you would add?
Emily: That was great. While you're doing the bootcamp, focus on getting through the projects – going from messy data to the results that you can present on. I remember some of the Metis instructors were good at reminding us to table some of those moments when you want to do more. You think that you need to bring in more data etc., but they told us, "Just get all the way through the pipeline, then you can go back and add stuff. Document the idea, write it down. It's valid. But you don't necessarily have to do it right now.” That focus on getting through a project cycle is really helpful because it prevents you from getting stuck in one area. And it's also a skill that is helpful in practice afterwards.
Wrapping ip 2019 and Making Predictions for 2020!
November was full of news about fundraises (and layoffs) and updates on past acquisitions!
This podcast episode covers multi-million dollar bootcamp fundraises, 8 new schools, and more!