Ask HN: Possible to pivot into ML research career without a PhD? | Hacker News Reader

Ask HN: Possible to pivot into ML research career without a PhD?

17 pointspurplerabbit17 comments

Hi HN. I have a career question.

My brother in law just started a masters program in ML, and is thinking of doing a PhD program after that. He wants to ultimately do ML research at an industry lab (like DeepMind or whatever). He’s not interested in academia, and is worried that he’ll burn out being a graduate student for the next 6ish years.

Would it be plausible for him to get a position as an applied scientist and then ultimately pivot into a research position over the course of a few years? Is that a viable path? Or is a PhD a must for this sort of position? I think he’d be happier working into that position as an IC rather than a student. He’s done some good research already, but doesn’t yet have published work.

View on HN

Realistically I think really high profile research labs will want you to have some published work or for whatever reason know that you are a really talented programmer and will hire you as a scientific programmer. I know at least one person that went to an industry research lab after a masters degree, but he had the necessary connections to there and the appropriate exceptional programming skills. Alternatively he could also consider doing an industry PhD, there should be lots of potential opportunities for that.

Appreciate the reply! I have a couple of followup questions…

First, what is an industry PhD? And second, do you have any perspective on the state of the field outside of the high-profile labs?

At least in Germany (where I am from), but I believe also other places in Europe, it is pretty common to do a PhD that is partially or fully funded by some company. For example Apple, Bosch was looking for ML PhD students at my university. Basically many larger companies, which might not come immediately to mind like Roche, Bosch, Merck, BMW, obscure Sawmill manufacturers, have some interest in Machine Learning. Only for a subset of them I would really call this research and as I said in most cases you either need to be exceptionally talented with connections or you are probably better off having a PhD.

I'm currently in this position. I'm a non-PhD (junior) researcher for AI and ML in Germany. I would say at least 50% of my research coworkers "only" hold a Master's and no PhD. However, I'm currently a bit uncertain about the future of my work and career. I'm not sure whether I should strive to "more", i.e. pursuing a PhD, however, it sometimes feels like a dead end not to do.

Hey there, I’m in Germany too! What domain or area are you working in? Any chance that you have some experience with image segmentation?

I'm working mostly on HCI (human-computer-interaction) related topics and natural language, so no experience on image segmentation.

Thank you for answering. I‘m currently looking for something like a „sparring partner“ to talk about what they are doing as well as what I am doing since where I work no one does anything related to Deep Learning or Image Segmentation unfortunately. Wish you the best!

> Would it be plausible for him to get a position as an applied scientist and then ultimately pivot into a research position over the course of a few years? Is that a viable path? Or is a PhD a must for this sort of position? I think he’d be happier working into that position as an IC rather than a student. He’s done some good research already, but doesn’t yet have published work.

This is _very_ rare. From my experience, all of my colleagues that are working in research positions (i.e., actively publishing to top-tier conferences) have PhDs. The most sure-fire way to work in research is through a PhD program at an R1 university because of the education, mentorship, and connections it provides. One option might be to start as a research engineer, collaborate with other scientists on papers, build a track record as a secondary author, and then apply to PhD programs, but that will still ultimately require getting a PhD.

This isn't meant to be discouraging, but I think that the most obvious choice for active publishing is through a PhD program.

If he wants to do ML research, why doesn’t he want to do a PhD program? I dont think you can get both: if you do research, there is less focus on making operational software. If you make operational software, there is less focus on doing research. Find out what you really want to do, and go with that.

It’s mostly the prospect of being a poor grad student for 6 years that’s disheartening

The various industry “AI Residency” programs were more or less created for this purpose.

I had honestly never heard of these programs. Thank you for drawing my attention to their existence!

What do you and your brother think people enrolled in PhD programs do?! Build paper planes and throw erasers at each other all day long?

Well, TAing? and making very little money… :)

If money is your focus, I'm not sure why you want to do research

I have been applying Deep Learning to novel scenarios since my last year of undergraduate degree and getting paid for it.

You could call that research.

I have applied DL to fields before anyone had done so.

My plus points were-

1. Being in a reputed uni (but nothing like an Ivy League or similar)

2. Completed a lot of MOOCs where I studied seriously.

3. Very unique and non-trivial projects (Titanic and House-Price Prediction not even on my resume).

4. Accepting less pay because cost-of-living is low in my country.

5. You could say that I am a good communicator (I have been in performing arts and writing since single-digit age. These do give you advantages.)

I just finished my undergrad few months back. I am still with my old company. But will join a serverless computing start-up this December by a very successful serial entrepreneur.

This work will involve applying Self Supervised Learning to Vision problems.

Now, none of these "work" is full-time, per se.

I never accept full-time employment where you have to attend stand-ups and get micromanaged.

I am in no dire need of money.

I get approached by recruiters regularly.

The best one was from Aleph Alpha.

But I don't want to leave my studies at this stage, because one day, I want to work at bleeding-edge problems at Tesla, Google X, Microsoft Research, DeepMind, OpenAI, or similar companies of the future.

I don't want to do consulting all my life where I solve business problems with science and technology developed by others.

This is really interesting work, and I acknowledge that. But still, I want to work at somewhere where my work can impact a lot of lives.

Now, is it possible to do good research without a PhD, as in "ability"? Yes, absolutely.

But good luck with convincing that to the people that will matter.

The big companies, and Elon, personally push open education, does not push university education publicly.

But please look at the people they hire.

Attend any public, or not so public Machine Learning event, and look at the profiles of the speakers. They all hold degrees from very reputed unis (Stanford, Berkeley, and like that), and Masters and PhDs at that.

They will only hire those people. Whether it is DeepMind or Hugging Face.

Now, despite my thinking that I, personally can have a good career without a Master degree or a PhD, I am not risking it.

I am getting into a Master's this year or the next.

I would also like to do a PhD not only because I want to work at DeepMind some day, but also:

1. I genuinely enjoy teaching and would like to remain in a regular teaching profession- full time or part-time. Having a PhD makes this easier.

2. There is always the attraction of pushing the knowledge of the human race.

_____

To answer your question directly, I would say, yes, it is possible, but your brother would have to be really extra-ordinary. It is possible to find people who have graduate degrees from Stanford and are extremely capable and are interesting people. And if your brother does not have the former, he must have a lot in the latter.

And still, it will be hard to convince the people that will matter.

I will focus on-

1. First and foremost, get really good and original papers out in journals, and not just arXiv.

2. Do non-trivial projects that shows capability and originality in thinking.

3. Network really well.

4. Grow soft skills like writing, communication, etc.

You have to publish something first.

Reply on news.ycombinator.com