14 December 2015 | by Piotr Migdał | 5 min read
tl;dr After finishing my PhD in quantum information, I turned to data science. I couldn't be happier about this transition! Compared with academia, data science world looks to me like a wonderland.
When meeting acquaintances, the typical conversation goes as follows:
I do data science freelancing. That is, I take contracts related to machine learning (predicting things, e.g. user growth of a company), data visualization (custom charts in D3.js), preparing and conducting trainings in data analysis (e.g. in scikit-learn and Spark), and other data expertise tasks (e.g. merging data from various source). It's rather a consulting-like service than being a code-monkey.
I aim at short projects - I like energy, deadlines (sic!), and novelty. Typically I run 2-3 projects simultaneously. Some projects are purely commercial, some are academic or for NGOs (but still with grant money), some are my personal side-projects. In almost all of them I have a lot of freedom in setting how I am going to solve a problem. In fact, usually I am required to propose a solution to their problem, e.g. to design a data visualization, not only code it.
Right now, I love it! Here is why:
I feel that the programming world is much more meritocratic than academia:
Much more $; while I was never too greedy, it has a direct consequence:
The overall difference is tremendous - from swings between depression and frustration (with occasional rays of hope) and a very unproductive state by default, to feeling great and learning a lot. Of course, it is also a function on my personal traits (chaotic, hot-headed, multidisciplinary, loving data, disliking hierarchy, a bit narcissistic¹) and my particular experience of academia.
I was afraid that the escape from academia would burn bridges, especially as I was openly saying I wanted to move out straight after my PhD defense². Yet, from time to time I am being invited for scientific talks or collaboration, even when I make it explicit that I am no longer in academia (or aim to be again). Furthermore, I feel that it's easier for me to contribute to education as an independent freelancer than from within the system³.
Very often I am being asked by academics how to make the transition from physics or mathematics to data science or programming. I do my best to give as detailed, personal and practical answers as possible. When I get such emails, it's flattering when it comes from undergraduate students, I feel sympathetic when it's from PhD students, and it is sad - when from people, who have invested in academia a number of years after their PhDs. Especially ones that love academia with its quirks, but have family to support and cannot bear the lack of security and constant move.
I have never regretted my transition. Instead of asking myself "why I am doing this" I am happy.
Don't try to treat it as an unconditional recommendation of data science freelancing!
Data science is great, but you need to enjoy programming, dealing with real & dirty data and with "good enough" solutions. (But no, you don't need to work in adverts or finance.)
Freelancing is a perpendicular dimension (the "normal way" to do data science is via full-time works). Whether you love or hate freelancing boils down to your psychological traits and whether you are able to get clients. As I like to say:
Freelancing is something exactly between having holidays and deadlines all the time.
There are a few follow-ups I'm considering:
How to move from academia to data science? (mostly: practical links)
If you like to hear about one of these topics, mail me! :)
I would like to thank Michał Kotowski, Marta Czarnocka-Cieciura and Jacek Migdał for comments on the draft. This blog post started as an email to Artem Kaznatcheev entitled "what I do" I wrote when flying from Dublin to NYC. Credit for the doge PhD picture goes to Ryszard Paweł Kostecki.