I think everything boils down to finding an actual demand for your skill set. As long as some organization or an investor has a real problem to solve and you co

I think everything boils down to finding an actual demand for your skill set. As long as some organization or an investor has a real problem to solve and you could convince them that you really have the ability and necessary training and experience, then it doesn't really matter how do you work and from where. You could always frame a contract not for hourly basis but on completion of the prototype, milestones, whatever you like. Basically, you should imagine yourself as a specialized company doing sub-contracting in ML. Finding a real demand and securing a contract is much harder than doing actual coding and modeling. There are literally thousands of guys who are smart enough to apply ML, especially now, when it became mainstream. PhD is, of course, a huge advantage in the world of pointy haired bosses, HR and paper pushers. With it you, probably, are looking for a very high hourly rate instead of subcontracts. In this you probably would fail, because people who are willing to pay high hourly wages needed certain rituals to be performed publicly in return.