Can an applied (e.g. applied computer-science) postdoc compensate for pure PhD (e.g. Pure Mathematics)

Question

I have graduated last year from a PhD in pure mathematics and I intended to do some machine learning in industry. However, seems like no place takes me seriously so I began asking colleagues from industry why. They claim I am on the one hand overqualified to do a "usual" programming job but on the other hand under-qualified to do a "serious algorithms job" as I have never worked in my PhD on applied computer science/Machine Learning.

One colleague from a technology company said

Doing a PhD in pure math was a sin. Repent for your sins by doing a post-doctoral degree in applied Machine Learning.

What do you think about this idea of doing applied postdoc just in sake of "cleansing" a pure PhD from the cv? I feel deeply traumatized from the academic poisonous environment I had to suffer during my PhD training and I don't think it'd be any different committing more to the same evil. On the other hand, can it really help improve my value in industry ? Let alone when applying for bigger companies?

Answer 1

Since I have a pure math PhD but worked in CS (academia) throughout my career, I don't see any need to "expunge" a degree. You are in a good place to learn computer algorithms in a short time since you have the background on which they are based.

The problem would only be gaining some recognition in a different field. I've also done that. Plan to attend some (many) conferences and get to know people. Industry folks go to these as well as academics. Try to form collaborative relationships and work (with others) on a few papers. It would (IMO) be faster and easier than taking an additional degree, much of which would bore you, I suspect.

Answer 2

The key issue here is that your qualifications do not match employers' requirements if you are looking for AI jobs. They seek (and find!) employees with a software and algorithmic background, and for all of the hard work that went into your PhD (of which you should be proud!), you just don't have those qualifications. It is, then, not a surprise that your applications do not lead anywhere: You would also not be hired as a sociologist for a think tank, or as a welder by a steel manufacturer, PhD be damned! Having a PhD does not make you qualified for jobs outside your area of expertise.

The way forward then is to either (i) work on having the qualifications the employers for your dream jobs seek, and being able to document them, or (ii) change your target for jobs to things you actually are qualified for already.

Answer 3

I can understand why a company would be worried about hiring a pure math PhD who doesn't have applied experience. I have known a lot of mathematicians, engineers, and people who have migrated in one direction or the other. It's not a guarantee that you can implement something -- or even develop something that CAN be implemented -- just because you can write a mathematical proof about it.

So: The piece that appears to be missing from your resume in order to get a serious algorithms job is proof that you can take things from the conceptual high-level pure math domain and implement them in code. If you want those jobs, the question you should ask yourself is: How can I fill that gap?

A postdoc is an obvious way to do that, which is why your colleague suggested it. You will be hired to do some specific implementation work which will result in products (papers) that you can point to, and you'll have someone who can be a reference and vouch for your ability to make those connections. That puts you in a very strong position.

If you are traumatized enough by your experience as a PhD student that even spending one or two more years in academia feels unmanageable, break down those two benefits and look for other ways to get them. You could implement some aspect of your PhD work in a code base that you put on GitHub. You could write about that and put it on ArXiv. You could reach out and make contacts with people who you think might have applications that your code base would be useful for, and ask if they would be interested in collaborating.

Just the first thing -- the GitHub repo -- might be enough to get you hired. If it's not, each subsequent step will put you in a stronger position. But each of those steps takes time. Again, this is why a postdoc is handy: You get paid.

The other option you mentioned was "usual" programming jobs. You say you are overqualified, but you also didn't mention your level of expertise in software engineering. Could it be that you are underqualified in that regard? Again, GitHub repositories will help you if you have that problem. You can also take one of those coding boot camps, but it is much harder to get placed with a company through one of those programs these days than it used to be. There are programs aimed specifically at people who did a little programming in their non-CS PhD program and now want an industry job that uses a lot of programming that might be helpful, though.

If you are already extremely proficient in programming, so the only problem is that you are overqualified, I will remind you that you can list your PhD as work experience rather than education. If you focus on the transferable skills and minimize the cachet you expect a PhD to have, it will probably read less to a hiring manager that you are overqualified. But I suspect, given your background, that it's more likely that you are simultaneously too expensive to hire (because of the PhD) while also being underqualified in other important areas (ie with practical experience collaborating on large code bases) for the lower-level software engineering jobs.

From the perspective of a hiring manager, the PhD can mean you will require a higher salary due to company policy, or because they expect you to negotiate harder because you have the degree. The PhD can also mean that you'll ditch this job as soon as you get one where you can use the PhD, which is expensive because replacing and training people takes a lot of time. I haven't been on a hiring committee for machine learning jobs in industry but I've had an inside look at several hiring processes over the years and the jobs tended to go to someone who would grow into the role rather than the most highly qualified candidate because they were expected to want to stay longer in the position. If you're really good -- much better than the other people applying -- the hiring manager is more likely to be willing to take that risk. If you have an equivalent level of practical experience with software engineering to someone without the PhD, they may not be willing to take the risk. (Also, please understand that not all hiring managers will feel this way -- a minority will view a PhD as a plus no matter what.)

I hope some of this is helpful for you.

Answer 4

This quote seems like a silly attempt at a joke, combined with a serious attempt at gatekeeping. The majority of people working "senior" AI / dev jobs do not have any serious background in maths, be it pure or applied, and won't be competitive with a candidate with Pure Maths PhD. They have some advantage by "knowing the ropes", i.e. the processes in their current workplace, libraries, people one need to please to pass a review, some tricks of the trade, etc. But doing a PhD prepared you intellectually to do serious research at a much more rigorous and detailed level.

Don't let anyone tell you that working hard for your education was anything other than a noble and difficult act. Don't let anyone tell you that you don't deserve your dream position because your educational profile is different from theirs. You have a terminal degree, there is no degree higher than a PhD. Keep applying to top companies and eventually your achievements will be noticed by people who value and recognise them. Good luck.

Answer 5

idea of doing applied postdoc just in sake of "cleansing" a pure PhD from the cv

I work in the industry and hire people. I do not know what a "postdoc" is.

I see on your CV that you have a PhD in "mathematics", we want to hire for "machine learning". This does not really match.

I look at your experience in machine learning and I do not see anything.

Next.

---

You may be more lucky but it is likely to be the way your CV is perceived.

The only reasonable way to get out of this is to show that you can do machine learning. Depending on where you are, the industry and the alignment of planets, this may be

• a degree that says "machine learning" (or "data science" - close enough). The fact that it says "postdoc" does not change anything, it could very well be a new degree from scratch.
• a serious set of things that show that you have experience in the matter (projects, etc.).

Please keep in mind that the path your CV takes into a company is of uttermost importance. If you apply on the web site your chances are slim. If the CV goes through someone inside the company who can advertise your added value then the chances are much higher.

Your degree in math can be very valuable to companies that take machine learning seriously, to the point of enhancing the mathematics behind (and not just throwing data and hoping for the best). You need to find the companies that are in the first case and then apply to the people who work on the enhancements.

Answer 6

There are no real post-doctoral degrees, but I agree with your colleague, though not with the facetiousness:

You have shown a great collection of skills by getting a degree in Mathematics.

An employer might worry about getting you up to speed on ML, which has considerably evolved in the last years. People in your situation that are doing research work have progressed quite a bit, including in the personal skills that they now possess. Doing a Post-doc in ML would be great to overcome this problem. In a related manner, writing papers in ML would also help.

There is no need to "expunge" your doctorate. If you can prove that you acquired ML skills, your degree gives you some advantages over people who do not have the degree of abstraction, intuition, and precision that you acquired. Others in ML of course have acquired those through a different road.

Answer 7

Your sample is non-representative. I have at least 1.5 acquaintances who have a theoretical math Ph.D. and post-doctoral work - category theory - and have gotten work doing on some some deep-learning-AI-thing, or cryptography.

Also, you are not over-qualified for a usual programming job - you're under-qualified for it: A "code-monkey" requires knowledge and experience with programming languages and development tools more than they need deep abstract insights which a Math Ph.D. affords them.

What do you think about this idea of doing applied postdoc just in sake of "cleansing" a pure PhD from the cv?

You're thinking like an academic. Don't do that. Take some time to:

• Read up on ML classically (maybe this) or the fashionable deep-learning-by-neural-networks stuff.
• Study and practice a programming languages (Python is quite popular where it comes to ML), perhaps contributing to some relevant FOSS project if you find one that interests.

As you do this, you can add relevant items to your resume; and in an interview, you could explain how you are making your way into the field. Also,

• Try to find some "meetup"-like activities in your area regarding ML or the sector of industry you want to get into, and attend