Why did you decide to no longer pursue an academic career?
Like many people, a career in academia looked like a great career choice when we started, but we slowly realized that it wasn’t for us over the course of the multiyear experience. There isn’t a lot of career flexibility, the publication cycles are very slow, and the problems we work on are often trivial compared to the impact we can have in industry. So transitioning into data science seemed like a natural role.
How does the research you do as a data scientist differ from an academic position?
There are some commonalities. As a part of the data science team, I leverage my PhD skills of interpreting and understanding data to build data-driven products and offerings that would delight our users. However, research as a data scientist is much more exciting. We work on products that are used by millions every day, rather than papers with a narrow academic circulation. Ideas we dreamed up in the morning could be deployed in the afternoon and we would see the impact of our work immediately. This is a far cry from the multi-month and sometimes multi-year publication cycles of the academy.
How does having a PhD help you as a data scientist?
While you don’t need a PhD to do data science, the degree certainly helps. My degree was in applied mathematics, and a rigorous, thorough understanding of probability and statistics was extremely helpful. My research also gave me a lot of experience with handling data. Having direct experience handling messy, real-world data through my PhD is a skill that certainly helped me get a data science role and helps me every day in my job.
What advice would you give others looking at an alternative career in data science?
My experience has been that the academy really prizes theoretical, rather than practical, skills. Nowhere is this more evident than the attitude it places on programming and data analysis outside of computer science. Professors often outsource the “lowly” job of doing the analysis and programming to graduate students, with surprisingly little supervision. This is often because the professors themselves don’t have the requisite background to teach these skills. If you’re a PhD looking to get out of academia, seize the opportunity to get as much programming experience as possible. Our blog on the basics of computation is a good place to start.
What is the biggest obstacle you’ll face in getting a job as a data scientist?
The biggest obstacle you’ll face is the stereotypes about the poor quality of academic code. Professional data scientists and software engineers have a strong aversion to poorly structured code, so set yourself apart. If you’re still writing long, single-file scripts with comments to set configuration runs, consider improving your programming style and showcasing it by putting your code up on GitHub.
How did your network and/or personal brand help you get your non-academic position?
Get out of the academic bubble. Go to meetups and get to learn about the industry. What kinds of data scientists are there? Which one are you? Take classes in data science. Build up your personal github account to demonstrate that you have coding chops. Meet people from programs or schools who’ve made the same transition. Take classes on programming tools or data science to brush up on your data skills. If you’re an international student, become familiar with the visa rules that apply to data scientists. Learn how to relate your field to society more broadly like education policy, polling error, algorithmic bias, or 17th-century naval technology.
How should you learn to market your PhD skills for a data science role?
I actually think that all STEM PhDs are helpful for a career like data science, but PhDs tend to be bad at marketing themselves. One of the most painful things to watch is seeing extremely talented individuals stumble over an answer when asked about their skills or how their PhD relates to potential employment. This is a reasonable ask, especially since half of data science is telling good stories. And it’s part of a necessary wider mindset shift in transitioning into industry, whether for data science or not.
A common compensation mechanism is to resort to general non-falsifiable platitudes that could apply to anyone (“I’m a quick learner,” “I’m very independent,” “I’m very hardworking”). These cliches may or may not be true of PhDs. They might even be true for you. But if that’s the best skill you can offer after five-plus years of research, then you’re going to have a problem.
The problem is that in academia, people stick to their niche and end up speaking only to others in their discipline. And so there’s no need to relate your work to a broader audience or adapt to ever-evolving audiences. When transitioning to industry, you need to explain to a potential employer why what you did was interesting, and how it will help you do the job. At The Data Incubator, we run a fellowship that has helped thousands of PhD fellows transition from academia to industry. We tell our fellows to figure out what specific skills the manager is looking for and weaving a narrative of their PhD work that targets each of those skills. Of course, we teach industry-specific skills like Python and Spark, but they combine that knowledge with their research background to highlight their deep expertise with data and statistics. But all of this has to be delivered quickly — within 3 sentences or less. You want to incorporate this into your resume. It’s not easy to learn and requires a lot of practice but it is highly effective.