I'd say you have a big advantage over proper postdocs because you have the salary that allows you to build your own 4x3090 training server. Combine that with a

I'd say you have a big advantage over proper postdocs because you have the salary that allows you to build your own 4x3090 training server. Combine that with a multi GPU Jupyter docker image and you'll be able to do experiments over night where most other researchers need to schedule cluster time and wait a few days. Also, you will most likely have a lot of experience with tools and solutions that aren't known in the research world, so you'll have an edge in out of the box problem solving. And lastly, it'll be easy for you to come up with research topics because you know what solutions were missing at work. Plus it'll be easy to get industry buy-in for those problems, which is mandatory for some grant money allocations. Now you just start some interesting research as a hobby and when you have the first results you reach out to people at your old university and publish a paper together with them. And just like that you're an active researcher again :) BTW I've been doing AI research for fun myself and so far I've "won" Sintel Clean (optical flow) and German speech recognition. So it's doable, don't worry :) If you want to reach out, my TEVR paper has the email address.