Thursday, January 19, 2017

Tyler Otto on Finishing your PhD in Physics and Going into Data Science

Tyler Otto is a data scientist at Reddit with a PhD in experimental 
condensed matter physics completed at UC Davis in 2013. 
Tyler Otto completed his PhD in Physics at UC Davis in 2013 with experimental condensed matter physicist Professor Dong Yu. Formerly the head of data science at Hipmunk, Tyler is now a data scientist at Reddit. He lives in Davis, has spoken with our students in our alumni seminar series, and is generally interested in contributing to his alma mater. 

Recently Tyler, Professor Ethan Anderes (of UC Davis statistics), and I met with six of our current graduate students in physics to hear about their research projects as we looked for ways we could potentially help with the data analysis challenges they face. After the meeting, Tyler realized one way he would be happy to help out would be to offer his thoughts about how to transition from doing a PhD in physics to working in the field of Data Science. The rest of this post is the entirety of the resulting email from Tyler to the students. 

I want to start with some of the concerns that people have about freshly minted PhDs.

1.     Overly academic: As a PhD student, you worked on projects that took years to complete, and you were driven by trying to understand the most challenging phenomena. Contrast this with what most businesses need, quick insights that can be acted on immediately. Most companies are not at the level of sophistication that your respective field is (your field of science has likely been around centuries, far longer than most businesses). In physics terms, most businesses are still learning their kinematics equations, but many new candidates talk about Quantum Mechanics.
2.     Communication skills: You have been surrounded by people that speak your same language (science, I mean), and are generally deeply knowledgeable of the types of things you are working on. Similarly, when giving a talk, you are used to presenting to people that want to understand every detail of what you have done and why you have done it. As a result, academic researchers often develop a communication style that is very different than a typical person in industry. As a Data Scientist you will work closely with marketing teams, engineers, business development, VPs, etc. These people are all very smart, but require that you learn to adapt your communication style.
3.     Too specialized: This is the easiest problem to address. You spent the last 6ish years thinking about a particular set of problems, and likely addressing them with a handful of tools. For many of us, these tools did not include lots of programming, machine learning, statistics, etc. How the hell are those hundreds of hours in the clean room, or in front of an AFM going to make you a qualified data scientist!?!

Let’s address ways that you can convince potential employers that their concerns do not apply to you

1.     Focus on impact: Many inexperienced candidates tend to talk a lot about tools, and interesting problems. Instead, try to identify potential problems that the company may face, and explain what you can do to make an immediate impact on this problem (this also shows that you did your research and are somewhat knowledgeable of the company). I also suggest showing that you can scale up your complexity. “Here are some simple ways I would initially approach this problem for quick results. If we see the impact we are hoping for, investing time in a more complicated solution such as (insert buzzword)”
2.     Have an elevator pitch: You should be able to explain, in 30 seconds or less, what your PhD research was, why it was important, and what contribution you made to the field. This sounds’s not! Some people may ask you to go into some more details, this is not an excuse to to show off how smart you are. Spend a few minutes (or less) touching on some of the broad strokes of your research. If you did something cool, like work at CERN, feel free to mention that (not in a bragging sort of way). This is not a conversation with a fellow researcher, this is a cocktail conversation. I believe this is the hardest problem to address on this list. We have developed our communication style over years, it is going to take a lot of work to undo some of these habits. On top of being the hardest thing to change, it is the biggest reasons that I reject candidates. START PRACTICING NOW!!!
3.     Identify your core competencies: No one is hiring a recent PhD in physics because they are a machine learning expert (unless you are). They are hiring you because you are smart, have excellent critical thinking skills, have worked in collaborative environments, learned to ask hard questions (and answer them), and more. Convince yourself of this and be able to articulate it to a potential employer.
4.     Learn the tools (but don’t be a cliché): This is what most people focus on the most, but it is the most easily solved of the problems that new candidates face. Inexperienced candidates tend to throw around a lot of buzz words (neural network, machine learning, collaborative filtering, etc). This is the equivalent of interviewing a contractor based on their discussion of hammers, drills, saws, and levels. This should not be the focus of your discussion, but it is important to indicate that you have some experience with these tools. An exploding number of online resources has made it trivial to gain some basic facility with the tools you will be expected to use. There are tons of other sources sharing which skills you need as a data scientist. Do a bit of research and focus on the ones that come up repeatedly. Focus on the basics before getting more sophisticated. You will likely do more basic statistics than you will machine learning. You will certainly do more basic coding than anything else.

Finally, I would like to share my thoughts on how you should proceed from here and what you should expect:

1.     Start figuring out how to apply data science tools in your research. It will make for a better conversation during your application, and will likely improve your research.
2.     Start applying for jobs right away. It will give you an idea of what the application process looks like, and it will help you tailor what you focus on. It will also help you work on your communication skills. It will also help you optimize each of the hiring process.
3.     Work with your career center. They will help you with your interview skills, resume, etc...this is an excellent resource!
4.     Expect it to take 6 months or more to land your first job. Don’t get discouraged.

5.     Go to different meetups. They are boring, but they are a great way to meet people in the field you are interested in, and a good way to continue to develop your language.