Pivoting from Political Science to Data Science
David Gelman is a data scientist at Pivotal. Despite having two Master’s degrees and a PhD, the road to landing his first job as a professional data scientist was not straightforward. In this first-person account, David shares his story, including the setbacks and triumphs he experienced along the way, with the hope that it will help fellow data scientists with unconventional backgrounds as they seek to start their own careers.
I faced a conundrum entering the fifth and final year of funding for my political science PhD. The problem was that, for a variety of reasons — some personal and some professional — I didn’t see a future for myself in academia. Plan B would logically be political consulting or working with a polling firm, two roles for which I had little enthusiasm as they would necessitate a much closer relationship with politics than I cared to have.
I thought about my strengths outside of my subject area expertise in political science (more specifically: American political institutions). I possessed highly developed analytical reasoning, experience in high level data-driven research, formal training in statistics and econometrics, and competency in an object-oriented programming language (R). I was aware of data science as an emerging career path, but was apprehensive that it could work for someone like me with a background in a social science that wasn’t economics.
What interested me about data science was actually the same thing that drew me to academia: using data to answer questions. Wanting to know more, I began by learning from people who knew more about data science than I did. I read posts about the tools used by data scientists. I contacted friends, acquaintances, and colleagues of friends and family who worked as or with data scientists, and asked them to speak with me about a potential move from academia into data science. I am exceedingly grateful and indebted to all those who took the time to share their perspectives with me.
What I found was that there was a lot of information out there on what it took to be a data scientist, and how someone like me might successfully transition into the field. Unfortunately, what I learned was often contradictory. Someone would tell me that my experience with R was a strong enough coding background for potential employers to be interested and that I’d have no trouble landing a job. Others noted that without any experience with SQL and a lack of Python, I needed to set my sights lower on data analyst positions or at best, a junior data scientist role. The one unifying piece of advice I received in all my conversations was the importance of crafting a narrative for my journey.
I knew that I needed to make myself stand out and that the best way to do that would be to position myself as someone with solid fundamentals in statistics and advanced research with supplementary technical skills.
Only a handful of the many people I spoke to suggested a data science bootcamp. Admittedly, I was strongly predisposed against the idea for two reasons. First and foremost, the cost. Most immersive bootcamps in NYC cost at least $16K, though there are some exceptions. The second issue was time. The 12-week bootcamp and the estimated 90 days on average it takes graduates to get hired after the program ends meant I was looking at committing to up to six months of unemployment. Having just been in graduate school for seven years, I couldn’t stomach the idea of any more school.
Summer in the City
Stop me if you’ve heard this one before, but job hunting in NYC is tough. Despite all the demand for data scientists, hiring managers seem to value practical experience above all else. The longer I searched, the more obvious it became that even with my PhD in hand, companies didn’t seem to be interested in what I was offering. My frustration grew as the summer wore on, confident that I could excel if I could convince someone to focus on my strengths rather than my lack of previous experience. As I continued to try and improve my skills such as SQL and Python on my own, I also continued to network and reach out to whomever I could to learn more about how to better market my skills and position myself.
Towards the end of the summer, my networking paid off, and I began to interview for a data analyst position at a technology company. After a phone screen and a take-home technical assessment, I was invited in for an onsite interview. In response to my request for feedback when informed that I hadn’t gotten the job, I was told that I had fallen just short on the technical side of the interview and that there were concerns over my lack of real world experience.
I felt the message was clear. While I couldn’t control my lack of job experience, I could do something about my technical skills (or at least the kind of technical abilities that would be demonstrable during an interview). It was time to rethink my opposition to a bootcamp.
Once I decided to attend a bootcamp, I had to several options to choose from. In the end, I chose General Assembly (GA) for several different reasons. I felt its curriculum was a slightly better fit for my needs than other programs I evaluated; I liked the structure of its continuing career services (which would last until I got a job, however long that took); and I liked that it was a well-known name with a large alumni community.
To Bootcamp and Beyond
After only a few weeks in my General Assembly Data Science Immersive course, I started to regain my confidence. I enjoyed throwing myself into learning a new language and new approaches to modeling. Perhaps the hardest part of learning about data science was the change in the nature of the modeling enterprise. As an academic using regression analysis, typically all I would care about was the veracity of my coefficient estimates so I could draw conclusions about my hypotheses. As a data scientist, I was being asked to make predictive models — where the outcome of a single feature was far less important than the overall performance of the model itself. Both data science and political science required me to think critically about data relationships and understand underlying statistical concepts. However, whereas political science predominantly used to data to answer predetermined questions, data science was far more open to general investigation into a topic.
In early October I reached out to a friend who had mentioned seeing a data scientist role at the company where he worked (Pivotal). I was very interested in the role, had heard great things about the company from my friend, and read great reviews from Glassdoor. When he suggested I come to the daily company breakfast with him and meet some of the data scientists before I applied, I jumped at the chance. Breakfast was great, and I liked the vibe of the office. A couple weeks after applying, I got an email from a Pivotal recruiter asking to set up a call, which led to a phone interview with the hiring manager that was technical in nature, asking me a wide range of questions about different techniques and aspects of data science. I felt confident afterwards, but knew there had been questions I could have answered better. I looked at this as a great opportunity; each interview stage provided me with much needed practice and preparation for when I fully entered the job market at the end of my GA class.
I had my coding technical assessment the week of Thanksgiving. This consisted of a video call with a shared screen as I walked through the data analysis pipeline for a dataset followed by some SQL “whiteboarding.” The entire time I kept up a running dialogue of my intentions and thought processes behind my actions, determined to demonstrate care and deliberateness. While this is similar to the explanations I would provide in my research, the immediateness of the communication made it feel much different. I signed off of that call feeling very confident that I would move on to the next stage, which I knew would be in-person with members of the data science team. I also received some encouraging news during this time: a chapter of my dissertation had been accepted to a peer-reviewed journal had been accepted for publication! Since I had a PhD, the recruiter asked that I present something technical from my dissertation.
A few days before my job talk, the hiring manager gave me a call. She asked what I was planning on presenting. I told her I was going to present my paper (which used regression analysis but not predictive modeling) but planned to spend a good deal of time detailing how I would approach the problem from the perspective of predictive data science. She hesitated and then told me that it would be a lot better if I could present the results of that enterprise instead of just outlining it. I said okay — and then freaked out for about fifteen minutes — digesting the fact that I needed to redo a third of my dissertation analysis in the next forty hours.
After my initial anxiety, I realized that I had a choice. I had been told exactly what I needed to do to have a legitimate chance at an incredible opportunity, and it was up to me whether or not I would be willing to work for it. So I made my choice. I would do what I was being asked of me; GA had given me all the technical know-how I needed, and grad school had given me the ability to do clear and effective analysis and present my findings in a intelligible manner (often in a short period of time). These realizations motivated me to push past my doubts.
The next two days weren’t pretty, but I did it. I had a narrative running through my presentation, providing context for the project and doing the subject justice while providing enough technical detail and analysis to demonstrate my skills and abilities. Either way, it was a valuable experience, I now felt I had what it took to be a data scientist.
When the call with my offer came the following week, I was filled with immense joy and relief. The job offer gave me closure: my academic career was giving way to a new era. The best part? I only had two and a half weeks left of my GA course, and Pivotal wanted me to start after the new year. This gave me time to finish the course, and take two (much needed) weeks off to recharge before beginning my next stage.
Making the Pivot
I’ve been at Pivotal for a few months now, working on an exciting long-term project with the rest of the data science team. I’m getting an opportunity to learn new technical skills that I didn’t know when I started (Hive and Spark). While I am focused on one particular use case at the moment, it is great to know that going in this single direction now does not preclude me from going down a different path on my next engagement. The team members are kind and supportive, and I acknowledge, nearly daily, how lucky I am. Fundamentally, I think that I’m largely doing what I did as an academic: leveraging data to answer interesting questions, surrounded by curious and motivated individuals seeking to gain knowledge. I look back at the past eighteen months and see how far I’ve come, how much I’ve learned, and how much more there is to learn. It’s an exciting time, and I’m looking forward to what’s next.
Change is the only constant, so individuals, institutions, and businesses must be Built to Adapt. At Pivotal, we believe change should be expected, embraced, and incorporated continuously through development and innovation, because good software is never finished.