Objectively speaking, are STEM jobs more difficult to acquire than STEM graduate degree? [closed]

Question

I've recently been doing some (non-serious) job search. My background is STEM with a focus on computing.

What I have noticed is that the requirement to get a job with many tech companies seems to be much higher than the requirement to obtain a graduate degree and even after getting a PhD it does not seem that you would even match half of the requirements that these companies are demanding.

Does anyone else feel the same?

For example, I am currently looking at a company called Groq. For one of their few non-senior roles, it says: requires 2-10 years of experience in machine learning or software engineering, knowledge in hardware accelerators, familiarity with a subset of: linear algebra, Python, computer vision, natural language processing, C, reinforcement learning, FPGA, recommender system, C++, among others, and preferably publication in top-tier machine learning conferences.

I can firmly say that even PhDs whose primary research is in ML do not have these backgrounds that the jobs are looking for. This job description just doesn't seem to make sense in terms of what actually happens in school. A computer vision researcher is unlikely to be doing reinforcement learning and FPGA programming at the same time. A Python programmer is very unlikely to be somehow doing C programming at the same time and vice versa. Different people fundamentally uses different tools. Nobody designs integrated circuits while doing linear algebra (of any depth) at the same time.

This is just one job description out of hundreds I've seen and heard from other people. The numerous stories of STEM PhDs who are jobless or can't break out of academia seems to corroborate with my concern. I wonder if it is really true that it is harder to get a job nowadays than a graduate degree (or even a PhD) in STEM.

Can anyone who has been on both sides chime in on this?

Answer 1

Job requirements in IT are total bs and it's a well-known thing, and there are many kinds of it - small companies tend to post a magical unicorn requirements (be able to train NNs and write FPGA code at the same time) while big tech does whiteboard programming interviews which are utterly unrelated to the actual daily job process. Still, the job market is ever-shifting and somehow manages to handle all that.

The secret? In large, people in STEM-related industry use references instead of job postings to find jobs.

And to answer the actual question - yes and no. Academia does not prepare one for the industry job from day one as well as someone well versed in the industry can't just enter academia and thrive there. These are just different sets of skills.

In ML particularly, unless you're aiming for a "guru" consultant position, a single really successful model can become a business, so I'd say requirements are fairly similar: it's just how communicating results and requirements that differs.

If you ever consider taking the industry path - what matters is the now-acquired ability to learn and quickly sift through references (being able to do a cursory lit review in a day or less is a big boon!), job ethics, extracting valuable knowledge from experience and applying it to new problems... These things are universal, really, but one needs flexibility and agility rather than laser focus.

Both academia and industry like people who can solve problems, it's just that the latter has many smaller and commonly ill-defined ones whereas the former likes to go in-depth, as a general rule. And the one is not necessarily harder than the other.

Answer 2

Objectively speaking, are STEM jobs more difficult to acquire than STEM graduate degree?

Not in general. Individual jobs may sometimes be difficult to get into (for example, getting some jobs at Google or Facebook may indeed objectively be more difficult than acquiring a graduate degree), but most employment statistics I know agree that STEM graduates, to a large degree, get relevant jobs after graduation.

In that sense I am a bit suspicious about your claim:

numerous stories of STEM PhDs who are jobless or can't break out of academia

I have worked and taught in STEM for 15 years now, and from the top of my head I can't think of a colleague or fellow grad student that remained "jobless" for more than a few months after graduation. Not all of them scored great jobs, certainly, but everybody found something within the industry.

It's possible that the Machine Learning field is different since it definitely feels a bit overheated right now, but if you "can't break out of academia" with a decent ML degree you may be a bit too picky with what jobs you apply to.

Further, I agree with what Allure says in a comment. Your example job posting really doesn't contain anything that an applied ML candidate shouldn't have:

requires 2-10 years of experience in machine learning or software engineering, knowledge in hardware accelerators, familiarity with a subset of: linear algebra, Python, computer vision, natural language processing, C, reinforcement learning, FPGA, recommender system, C++, among others, and preferably publication in top-tier machine learning conferences.

Basically, if you did your PhD in computer vision, NLP, reinforcement learning, or recommender systems, you should have about 5 years of experience in ML, experience in at least one of the listed fields, probably quite some practical experience in Python and/or C++, and you probably have at least some experience working with hardware accelerants, plus some publications (if these are "top-tier" is always a subjective matter, I would not read much into this). All in all, you are in fact checking all the mandatory and even quite a few of the optional boxes. I see no reason why you could not apply to this job.