I'm actually going to suggest yes to you. OR at least you should start one. I, like you, cannot self-learn outside of work because my brain is done and needs to

I'm actually going to suggest yes to you. OR at least you should start one. I, like you, cannot self-learn outside of work because my brain is done and needs to recharge. I, like you, read plenty and do the courses, but don't actually do the problems. The real solution to your problem is to get another job. And I will disagree with some of the other posters that a PhD is required for ML because that is not necessarily true. I have a BS/MS in applied math and do research into how to apply ML into engineering automation problems for a fortune 50 manufacturing/eng company. Granted, we are not pushing the bounds of ML research. But we are looking at unsolved problems and using ML, control theory, robotics, etc to solve them. This is what you want. And yes it is very math/algorithms heavy. Whenever we try to hire more people, HR sends us tons of BS cs guys that are sorely lacking the math background. Most of our hires are engineers (mech/ee/comp/aero) or oddballs with math/physics/ computational science like me. So you can certainly look for other jobs (and if you want I will look at a resume to give more specific feedback). But if you already know that you don't have the background for controls and ML.. then go get it! Here's the thing with CS or compEng PhDs. You don't have to finish them to get tons of valuable knowledge and move into the 'cool jobs' out there. But (per the rest of this HN thread) if you happen to get an awesome advisor and team relationship then you might as well finish.