When I say academic-adjacent research in industry, I’m talking about industry jobs where your main responsibilities are to produce research that gets published. From what I’ve seen these are much higher paying than acedamia but have less flexibility. I’m mainly asking for the scope of ML research in big tech companies but I’m also curious to hear how this works in other fields.
I’m a newer student, but almost everyone senior in my lab does research internships with this objective in big tech companies, and most of them get return offers. Additionally, I meet a lot of people that have these jobs at the conferences that I go to; but then again this might be survivorship bias.
The issue is, most people I ask outside of the people in my lab or at conferences say this is extremely difficult or near impossible to achieve. I just wanted to gauge how realistic getting one of these jobs is? And how does the difficulty compare to getting a TT position in academia?
Comments
The aim of a business is to make products that are sellable for profit.
Publishing is what happens after the product is legally protected or the information is basically irrelevant.
There are no industry jobs where your main responsibilities are to produce published research.
It really depends on the industry and specific skills you have, but it’s generally easier to get than an R1 TT job. Companies seem cutting back right now though because of market volatility so it’s hard to say.
Research jobs at frontier labs in ML are very competitive. Hence the high salaries and prestige.
At my company we have universities do initial research on topics. Once they’ve proven the physics of it and that it is possible we take it in-house to apply to our products – frequently hiring those that did the research at the university to continue progressing it. We don’t really do much academic publishing.
On the company side, we have project engineers that manage the funding and POs to our university partners and discipline focals that help guide the university students.
In big companies there are research departments and development departments. The former is sort of what you are describing but the goal isn’t to publish its to do research that turns into a product or patent. Publishing is sort of a side goal and its rarely in true academic journals but they can participate in conferences and stuff. I wouldnt say its overly competitive just that companies are looking for specific skill sets
I worked 50/50 academics / industry for over a decade. We commercialised our research. It was a delicate balance but fun. A lot depends on your skills. We were developing tools for research instruments, so you sold the tools by showing they worked.
I’ll also add companies don’t just use publications as a means of advertising. In textbooks on managing industrial research there is a wealth of studies on compensation and incentive structures. Contrary to popular belief that bigger salary are always better, there’s usually diminishing return. It turns out that depending on factors like the stage of career, etc., a manager giving someone a vote of confidence on their work and public recognition by asking them to publish and present their work can have more of a positive impact on job satisfaction, career progression, etc. So publishing can also be because individual contributors and managers who rose through the ranks are also intrinsically motivated to publish.
In my subfield of physics, there are basically no industry-equivalent jobs. As in, if you use the top 5-10 technical buzzwords from my PhD research to conduct a job search, you get literally *zero* hits on linkedin, glassdoor, globally. Occasionally I’ll see one every few years, usually a fusion startup, and it’ll be a PhD-required job with only a few dozen applicants
I don’t know the academic side of ML, but in my casual looks at private sector jobs it seems that there are thousands of jobs with ML/MLE/DS/DL in either the job title or description. But each get hundreds to thousands of applicants ranging from BS to PhD holders.
Really hard to say for your field. For mine there is practically no job market outside of public research institutes like universities or national labs
I’ve seen multiple new PhD choose companies like OpenAI and Anthropic over assistant professor at Ivy League and top public schools.
They get paid 3-5x more than academia ($500K-1M total comp).