From astrophysicist to data scientist

People often sidle up to me at conferences, lab retreats, or receptions at the boss’s house.  “I’m thinking about leaving academia and going into industry/Big Pharma/law school.  What will that be like?”  Here’s a response to that question, not from me, but from a colleague who went from an academic career in astrophysics to a job as a research associate in a biomedical informatics department.  Here’s what that experience was like, both for him and for his spouse.  Since he wrote this, he’s moved into a faculty position.  His wife has gone into the private sector.  French vocabulary at the bottom of the post, as usual.

As far as the transition from astrophysics, I’m terrible at giving advice, but I’ll tell you the experience of my wife and I. My wife was a liberal arts college professor and I was on my second postdoc.  Both of us were working on high-profile cosmology experiments.  Solving the 2-body problem and the stresses of starting a new family were coming into conflict with our careers, and so we realized we had to make a change.  We had been doing astrophysics so long we literally had to mourn the loss of our future careers in the field – like we literally went through stages of grief.  Part of it was thinking about all the time we had invested and all the connections we had made, part of it was that we couldn’t imagine doing anything else, and part of it was that we were doubtful we had the skill set to compete with statisticians, data analysts and computer scientists who had been doing what they were doing since they were freshmen in college.

Eventually, we realized that the time in physics and in academia wasn’t wasted, that we had learned a lot about how people and organizations function, and how to get things done.   We also found that the physics data analysis methods and ways of approaching data were actually refreshing to people outside the field.  We further realized that people with statistics degrees know a lot about logistic regression, sample size calculations, statistical tests, .. but they are more clerics – they approach problems from the point of view of “I’ve spent years learning tests, which test best fits the problem?”.  This is in contrast to the approach of physicists who are expected to think of things from first principles and build things from scratch – scouring the literature for ways to solve problems, downloading random bits of code in whatever language and modifying it, …  This is a huge advantage, because people think we are brilliant when really we are just not doing what the usual statistician would do.  As far as computer science, yeah, we realized we weren’t programmers (my wife knows more about programming than anyone I know, but she’d probably have a hard time getting a job as a programmer at Amazon).  Further, in talking to people, we found that software certifications, in some circles, are actually taken as a joke.  So depending on where you’re applying, they may be more valuable, or less.  Obviously, it was good to show we had some knowledge of programming.  My only bit of advice: learn R – it is used everywhere.  It will literally take you 20 minutes and you can put it on your resume.  Eric Feigelson has some nice tutorials (e.g., https://www.google.com/?gws_rd=ssl#q=eric+feigelson+r+tutorial ).  Also, Hadoop and Spark are pretty much industry standards, so you might think about learning something about them.

We also found  many jobs BUT NOT ALL require you take ‘tests’ – almost like entrance exams filled with brain teasers and data science questions.  I could never do them… but almost all of the interviews required us to give talks (an advantage, given our backgrounds in academia).

Anyway, I eventually obtained a job in bioinformatics analyzing language and speech production of patients with mental illness at a teaching hospital, and my wife eventually landed in a company doing consumer data analysis.  The challenges are so absorbing, the only time I think about astrophysics is when  my boss asks me about some new astronomical discovery.

We were also worried about things like time-flexibility, especially since we have kids– like having to work a standard set of hours and put in for time off, but (1) a huge number of work places now allow you to spend a lot of time working from home (it’s almost expected in some industries), (2) many work places have flexible times when you can come and go, and (3) putting in for vacation has enormous advantages.  In astrophysics, I always felt like I had to be ‘on’ even during time off.  Vacation is a way of telling everyone, “I’m gone don’t bother me” and they are forced to respect it.  Also, speaking of the workplace, a lot of tech companies are still doing this ‘open office’ concept with no walls or anything.  It’s annoying and counter-productive, but they often compensate by having spaces where you can hide, and often allow you to work from home much of the time.

We were also worried that we would be working in a place that looked like the movie “Office Space” – with people with ties speaking in cliché BS bureau-speak.  There’s some of that (and it can be hilarious), but chances are you’re going to be working with other smart people who see through that stuff.

On a more superficial note, we also found that the word ‘astrophysicist’ carries a lot of weight.  I have had Harvard-trained neurologists (the super-brilliant nerds of the clinical world!) who always introduce me as “an astrophysicist” as if I were a brain surgeon.

Anyway, the bottom line is, we realized we were way more valuable than we thought we were.  Also, we realized we were looking for employers who were willing to take a chance with a person on a non-traditional background – i.e., non-cookie cutter companies run by people who understood that there would be a learning curve and respected our backgrounds.  We discovered that they are few in number, but very much exist.

  • la science des données ou la science de données: data science.
  • les mégadonnées: “Big Data.”  Le big data, littéralement « grosses données », ou mégadonnées (recommandé3), parfois appelées données massives4, désignent des ensembles de données qui deviennent tellement volumineux qu’ils en deviennent difficiles à travailler avec des outils classiques de gestion de base de données ou de gestion de l’information.
    (Source: Wikipedia.)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s