Corporate Data Science after a PhD in Physics

[Corpuscule]

I would like to start a thread discussing the opportunities for Physics PhDs in the field of corporate Data Science.

This emerging filed is briefly explained in the following article:
http://www-01.ibm.com/software/data/infosphere/data-scientist/

I have heard on a number of occasions that Data Science is a rapidly growing market, and employers in this area are currently willing to hire suitable candidates with Physics/Math PhD background and some programming experience due to the scarcity of qualified and experienced Data Scientists.

I will be updating the opening post as I research these opportunities further and come up with ways to increase the chances of breaking into the corporate Data Science field. So far I have found a few advices here: http://www.quora.com/Data-Science/How-can-a-physicist-get-into-data-science#

This is one of the posts from the link above:

1. If you have or are finishing a PhD, consider doing the Insight Data Science Fellows Program. This is essentially a networking/job-placement program that could make your career switch very easy, but getting a slot in this program is competitive.

2. If you can't get into (or don't want to do) the Insight Program, use the list of companies that they recruit for as a starting place for your job hunt. If your resume is less than stellar, look for positions at smaller and less prestigious companies. There is no shortage of data scientists applying to big name companies like Google or Facebook.

3. Do an (paid) internship. It's a lot easier to get your foot in the door for a few months than to get a full time job offer. The internship also shows that you have invested some actual effort into your career path.

4. Network. This one is pretty obvious, but the best way to get a job offer is through a referral. It's also a great way to find out about jobs you might not hear about otherwise.

5. If looking for a place to learn on the job doesn't work, you may need to actually learn the skills on your own to prove your programing and statistics chops. Take the coursera course on machine learning. Do well in a Kaggle competition. Make an interesting big data demo (link to it on Hacker News). Put some code on GitHub. Contribute to an open source project.

Please feel free to contribute to this thread if you (or somebody you know) have moved (or are in the process of transitioning) from Physics/Math PhD into Data Science.

 

[UltrafastPED]

I don't understand what you are asking for. It is a real field; they do like people with experience with large datasets, analytical skills, and programming skills. This set of requirements matches up well with some physics PhDs - but not all.

If your goal is how to find a job in this field - well, prepare a portfolio of your experience with programming, statistics, and analysis of large datasets.

 

[Corpuscule]

I don't understand what you are asking for. It is a real field; they do like people with experience with large datasets, analytical skills, and programming skills. This set of requirements matches up well with some physics PhDs - but not all.

If your goal is how to find a job in this field - well, prepare a portfolio of your experience with programming, statistics, and analysis of large datasets.

Basically I just wanted to ask people who are interested in transitioning or have successfully transitioned to share their stories. When people change careers, there are always more nuances and important details than just preparing a portfolio of experience with programming, statistics, and analysis of large datasets.

 

[Locrian]

I'll be interested to see what you learn or others have to add, Corpuscule.

I work with a lot of data in my job, but I think it's in a different sort of environment than the work you're looking at.

 

[UltrafastPED]

I think you would do better with this type of question on LinkedIn.

 

[Corpuscule]

I'll be interested to see what you learn or others have to add, Corpuscule.

I work with a lot of data in my job, but I think it's in a different sort of environment than the work you're looking at.

Could you please share what are you doing? Would your field of work be suitable for someone with entry-level Data Science skill set? I am just trying to hedge against potential oversupply of Data Scientists, now that Data Science is becoming increasingly popular :)

 

[jesse73]

This is a perfectly legitimate thread to have here. Physicist are finding work in this area so of course it would be useful to query for peoples experiences and advice.

 

[Locrian]

Could you please share what are you doing? Would your field of work be suitable for someone with entry-level Data Science skill set?

I work as an actuary. This means I'm a member of the professional organization, adhere to it's precepts and follow the framework established by the actuarial standards of practice.

In general, actuaries are responsible for estimating future contingent liabilities. They typically do this with models. However, models have to be constructed and calibrated, and this is done by examining data. I work with large data sets that contain membership, claims and financial data.

The tools actuaries use are unusual and most actuarial education in the US occurs in actuarial college classes (optional), preparation for the http://www.beanactuary.org/exams/, and on the job.

 

[ParticleGrl]

So I made the transition from a high energy theory phd (mostly pen and paper work with a smattering of fortran 77 programming). For me, it was a pretty difficult transition- after my phd my only postdoc offers were abroad, and for personal reasons I couldn't follow that.

Originally, I bounced into a job tending bar, and was pleasantly surprised to learn that if you work at the right place you can earn quite a bit more bar tending then as a postdoc, which took some of the pressure off of my very frustrated job search. I bought some statistic textbooks, and started playing with open source tools, and I started implementing somewhat non-standard methods on data sets I got from various kaggle competitions (www.kaggle.com). In the end, I lucked into my first job- a statistical society conference was happening in San Diego, and I was talking to a customer and somehow we started talking about various statistical problems, and he gave me his card. Turns out, he was in charge a small data-mining effort at an insurance company and I called and got an interview.

From there, its been pretty smooth sailing. After a few months with that employer, I got some interviews with larger companies, and took a job for a group that does consulting. Working for them, I worked with data sets ranging from 100s of millions of health insurance claims to a few hundred thousand inventory orders for a small convenience store.

My recommendations would be to find a data set (there are tons of things out there) and start doing something interesting with it. What you need, more than anything, is to be curious. The most important parts of the job are 1. being able to ask interesting (and useful) questions that the data can answer, and 2. being able to organize and structure the data so you can start to answer those questions. For everything else, tools exist (and where they don't exist, its easy to hack stuff together).

The good news is that while I sort of fell into my first job, both companies I have worked for would easily interview a phd with minimal experience. At the insurance company, you'd be expected to know some very basic information about how insurance works (i.e. "did you at least read about insurance on wikipedia before coming into this interview"), and then some questions along the lines of "how would you approach a data set that looked like ____?" If you put specific skills on your resume that overlapped with mine, I'd ask questions about them, but not "gotcha" type questions, just "do you understand the basics."

 

[Corpuscule]

Dear ParticleGrl and Locrian,

Thank you so much for your valuable feedback! So far I have started a couple of fun projects in R and Python to learn these languages and have signed for the Data Analysis course: https://www.coursera.org/course/dataanalysis

 

[Locrian]

Data scientists were mentioned in today’s SOA discussion panel “Will Big Data Transform Health Care?” When asked where to find appropriate personnel, including data scientists, both Eric Bokelberg of IBM and Carol McCall of GNS HealthCare stated that they were difficult to find.

Eric mentioned the IBM BigDataUniveristy and Hadoop. Carol mentioned exploring non-traditional channels to find employees.