This may not be the type of answer you're looking for but near me insurance companies (AmFam, Sentry, etc) pay extremely well for the area and pay big money to purchase startups
I and my colleagues have done consulting work for several insurance companies. I'd love to develop some sick claim processing/auditing software or data management solution for insurance, but it's next to impossible to get visibility into the data they're processing and where the improvement opportunities lie without working there or having an in.
I'm guessing the cycle is; work at an insurance company, figure out a solution to their problem, ask your boss if you can work on that solution, be told no, quit and start a startup to do it, sell it to that company, get bought by that company.
This sounds complicated but remember that there is little downside risk for the insurance company that buys you. If you don't make a good product, then you or your investors are out the cash, and they are out $0. This is probably 99% of cases. If you do make a good product, then they can just buy it later when the $ is worth less! (There is also a risk that a competitor buys you, I guess.)
> processing and where the improvement opportunities lie without working there or having an in.
This isn't institutionally crazy.
It's an old, old, story. Lots of people feel like they can come in from the outside to a complex domain, apply some "generic" techniques etc, and make changes with huge positive impact. They usually wrong, usually enough that many people feel safe just ignoring the possibility.
Chance of success is much higher by building capability internally for the techniques with people who already understand the domain well. This however runs into both internal politics and moribund institutions so can be a real challenge.
One of the big startup purchases that happened while I was there was of a company that used machine learning to estimate the value of items in a room, ostensibly to guess how much your furniture and electronics are worth. Insurance companies are already good at actuarial sciences so they don't need you to estimate how likely someone is to file a claim, they need you to do things they're not good at
I've done so much work in this space that I wrote up a patern that I modify as needed. The only data-specific issues you'll encounter are the underwriting rules - each insurer has their own secret sauce that they guard very closely.
