Roughly a year and a half ago, in the times when you could still sit at the same table with people who are not your immediate family, I was attending the annual meeting of the Belgian data science meetup organisers at Digityser in Brussels. It was an informal affair, with the aim to look back at the last year of events, share experiences and brainstorm a common strategy for future meetups. As with most meetings of this type in Belgium, this one ended with an opportunity to share a few beers with others and chat about the state of the data science field in Belgium.

After a couple of beers, I got into a discussion with Kris Peeters of Dataminded about the value (or the lack of) of strong academic profiles in IT. People who have met either of us know that we are both quite opinionated and are not shy to share our views, sometimes even a bit too strongly. During the discussion, I was trying to make an argument that PhDs in technical fields are very useful for an IT career, but Kris strongly disagreed with me.

The heated discussion left me thinking. I recalled that a couple of years prior, a recruiter suggested that I downplay the fact I hold a PhD when being interviewed. Conversations with several people who transitioned from academia to industry suggested that there is a bias against PhD holders in the private sector. I always felt that my academic career prepared me very well for the role of a data scientist, but perhaps Kris was correct in his criticism. Maybe my opinion was biased by my own singular experience?

Current situation in the EU higher education system and the IT labour market

In my search for answers, I was surprised by the lack of publicly available data on the attitudes of employers towards PhDs. My own attempt to survey the field also failed, with a handful of people providing answers and revealing only that people who hold a PhD believe that their degrees are valuable for an IT career, but not saying much about the attitudes towards PhDs in general.

I was able to find several EU reports on the topic though. The 2015 report on the need for STEM profiles on the job market confirmed Kris' concerns:

"There are also concerns about mismatches of a more qualitative nature, for example driven by a growing ICT intensity in the economy, by technology convergence, and changes in the type of tasks involved in STEM professions, which are not always reflected in the design of higher education programmes."^[1]

In translation: STEM curricula are failing to properly prepare people for careers in the industry. The report also cites the lack of work experience as a barrier of entry of STEM educated profiles into the commercial sector. This is true for PhDs as well, although it is somewhat absurd that years spent doing research and teaching are not counted as "work experience."

There is an apparent mismatch between the skills employers expect from STEM educated profiles and what the education system is actually providing. Here, it looks like Kris has a good point, and my own experience in the EU higher education system confirms some of these biases.

For example, some STEM research groups still follow a pyramidal organisational scheme, where the professor serves essentially as the CEO of the research group, is supported by a layer of post-docs who carry out the operational part of the work and delegate the work tasks to PhD and master students. In this scheme, the PhD student is focused on task execution, rarely talks to their PhD advisor and is often shielded from the "big picture" of the research project. They are usually not involved in writing research papers, but instead provide "contributions" to the results presented in the research papers. This kind of education system is naturally bound to produce profiles with a high degree of mismatch with the profiles needed in the IT industry, due to the lack of emphasis on development of soft and analytic thinking skills.

The pyramidal research structure is becoming less common, which is good, but the sooner it is completely eliminated from the higher education system, the better!

Demand for STEM profiles will significantly increase in the near future, as the EU economy is becoming more and more innovation oriented. The EU is aware of this problem and several initiatives have been implemented in order to improve the bridge between academia and industry. For instance, the Innovative Training Network programs financed by the Marie Curie Actions provide PhD students with opportunities of internships in the private sector in order to boost the transferability of their skills and the non-academic work experience.

My view on the value of a STEM PhD in data science

I always took it for granted that a PhD in particle physics prepared me well for a career in data science. I never really took into consideration that my academic experience might not be representative of the STEM PhD programs as a whole.

I completed my PhD in the US and spend 3 years in Israel as a post-doc before coming to Belgium for a second post-doc. Both US and Israel feature academic environments which emphasise critical thinking and strong communication skills on top of the technical aspects of education and research. I found this to be the case also at UCLouvain, where I did my second post-doc.

Europeans often describe American and Israeli academics as "aggressive", because they don't hesitate to make strong claims in presentations or challenge the speaker on any point. There is some truth in that. If you don't believe me, try to give a seminar talk at the Weizmann Institute and behold a first year master student publicly rip your ideas to pieces in front of a room full of people. It is a transformative experience to say the least, but one which can be of much use in the private sector: you have to be able to stand up for your ideas and your work. There are no people or companies who became successful by being insecure.

I spent a significant amount of time in my PhD career discussing particle physics with my advisor. During the hot midwestern summer months, these discussions would often take place on the front porch of his colonial style house, to a glass of cold lemonade or sparkling water (always a nice touch to get your brain going). I learned as much from these discussions as I did from the course work. Not only about physics, but about how to think critically. We would read preprints of recent particle physics papers and try to understand them, challenge them and find loopholes in their arguments. This was a way to identify opportunities for our own research directions. We did the same to our own work, in order to ensure that we are not missing anything important.

To this day, I still apply the same self-critical approach to all the data science work I do at B12, and I credit it all to my PhD education.

My experience with European higher education was that the European academic "style" is different (not worse, just different). Seminar talks in European universities tend to be more held back, with a lot less questions and interruptions. Students are a lot less likely to ask a question than, say, a professor In part, this is likely cultural, but it does come with a more pronounced feel of hierarchy, and I always felt that on average students lacked the confidence to stand up and defend their point of view. At UCLouvain, I was very happy that I was allowed and even encouraged by my advisor to explore my research ideas, no matter how crazy they may have been. However, my work and ideas have landed me in hot water with other faculty members a time or two. Most European universities nowadays have requirements for students to write papers and give oral presentations in order to complete their degree, but my impression is that the emphasis on these skills is a lot less pronounced in Europe than it is in the US for instance.

A PhD requires you to be able to learn quickly, and learning quickly is a skill you have to learn itself. There is usually so much material to ingest that you have to be able to absorb and understand fast. I did this by first developing a solid basis in mathematics and physics that I could easily build on. Finding commonalities between new subjects and the ones I was already familiar with helped me to learn the material quicker. The more I learned, the easier it was to add to it.

My PhD program also had a rule that each student had to give at least one seminar talk per semester. Most people did it more often. This meant that you had to stand in front of a room full of people and defend your ideas at least 10 times before you defended your thesis (not including conference and workshop talks). My PhD advisor also pushed me to personally write all of our papers. I am eternally thankful to him for this, even though at times it was a frustrating experience. All of the above allowed me to develop confidence in public speaking and the ability to communicate my ideas clearly and efficiently.

When I left academia to join B12, I felt that the skills I acquired during my PhD years prepared me well for a role of a data scientist: I had all the technical, analytic and communication skills I needed, strong math and physics background (yes, knowing a lot of physics is still very useful in my day to day job) and the ability to learn quickly. I still feel that at its foundation, data science is a research discipline.

The transition wasn't completely seamless though. I had to re-learn how to code as software development standards in academia are about 10 years behind the commercial IT sector.
Even though the way I approach data science problems and the way I attacked academic research problems is similar, I had to unlearn that in academia you essentially have an infinity of time to work on a problem. In the private sector, and especially in consulting, the resources for your project are limited, and you have to learn how and when to stop. The focus of your effort is also different. In academia, simply understanding a problem has value itself, even if you don't find a solution. In commercial data science, every project is an investment, and the result has to generate value (i.e. either make or save money) at some level.

My impression is that learning all the IT relevant skills I didn't have as a PhD student was quite possible to do on the job. However, I maintain that learning the skills that I developed during my PhD on the job is much more difficult outside of academia: analytic thinking, solid mathematics foundation, the ability to learn quickly... these skills you develop by being in the trenches of science for a few years.

Finally, I think it is fair to say that there is only one kind of person who completes a PhD: the kind that is genuinely curious about understanding how the world works. The kind that is persistent, yet flexible and creative. And it is precisely this combination of curiosity, flexibility and creativity that makes great data scientists.

Is a PhD in STM an asset for a career in data science?

My short answer to this question is: yes.

The long answer would go something like this:

It is clear that there is a mismatch between the PhD skill sets and what the industry expects, but overall, I think that most people with a STEM PhD can bridge this gap. The fact that there are several initiatives aimed at bridging industry and STEM academia at the European level gives me confidence that the situation will improve in the future.

Keep in mind that PhD is not about the degree, it's about the abilities you'll develop during your PhD. These include strong analytic and critical thinking skills, ability to learn quickly and strong communication skills, on top of all the technical skills that you get as a part of your STEM curriculum.

If you are considering a PhD program, you should make sure that you will get out of it with the above mentioned transferable skills. If your PhD boils down to someone giving you tasks to execute without asking you to think about them in the context of the big picture, if you spend your entire PhD without writing a paper of giving a talk, then your PhD program is failing you. This is relevant both if you continue your academic career and if you transition to industry.

At B12, our experience is that strong academic profiles, especially PhDs in STEM bring exceptional value to our team and define the scientific DNA of B12 as an innovation company.

We expect our data scientists to be able to function as science mercenaries: trained in guerrilla tactics of data warfare, able to adapt to any domain, and develop creative solutions to any data driven problem. STEM PhDs usually fit this role very well! Currently about 40% of our team holds a PhD, and we continue to actively seek STEM PhDs in the recruitment process.

This is not to say that you can not be a great data scientist if you don't have a PhD. About half of our data science team does not hold a PhD and they are all very good at their jobs! However, when they were recruited, they were recruited because of their analytic thinking skills and their ability to learn quickly above all else, and these are the skills that should come as a given with someone who holds a PhD in STEM.

1. ^[1]
2. Does the EU need more STEM graduates?: final report, 2015. ↩︎