Many universities have a tech transfer program, ask about those. Keep in mind anything you develop with their time and money, they have rights to, which are usually arranged as a license royalty.
So what I did was get involved with those programs, learned a lot, started something different after graduation
I would assume 95% of startups that use ml do it for stupid marketing and fundraising reasons.
I also think once you are done with your PhD the field will have changed in a way you simplycant forecast.
I suggest it's more important to find your interest and passion. That will enhance your. Chance of a great thesis and lead to better outcomes.