That work would be far removed from the industry forefront. With none of the resources and team with you. It takes millions to develop, train and run these AI models. Your plan sounds dreadfully lonely to me.
On your deathbed you are likely not going to look back favorably at all the years you put in for bigcorp advancing the industry. Even if you do have a few white papers published, so what? Effectively nobody will remember them in 10 years anyway. Trying to find meaning by working for others doesn't pan out for most people. Bank the cash, get out when you can.