Defintely start your own company. You spent enough time in the industry to know what problems are worth solving. Take a very specific problem and solve it better than any generic solutions available. Be the king of your narrow niche market, and when you're successful maybe broaden it.
My favourite stories of engineers bootstrapping their own business with no investors, no employees, no MBA:
People talk about "chances of success" as it was a lottery with binary output win (you are millionaire) vs lose (make no money and close).
Well this is false:
- Define success: these tiny one-man digital businesses ran from home have so little expenses that going profitable is infinitely easier than any traditional VC funded startups with offices and teams.
- Speed: this is counterintuitive, but true: a one man band can take decisions and, pivot, optimize, and iterate at least twice faster than any team (no meetings, no presentations, no democracy).
- Time: when you are profitable, maybe you're not rich yet, but you have time. Time to add value to the product day after day. And you will get better and better at it.
If you create a subscription based business, even if you fail for the whole year at improving your conversion rates (you won't), your yearly recurring revenue will grow linearly. Meantime, your expenses are still minuscule.
Yes I think so! The pattern is this: big organisations produce generic solutions to cater 80% of the needs of the 100M people. Your task is to create very specific, expensive solutions for ~500 people.
Not a compilers expert, but I will try to invent a fantasy example:
GCC/clang are ok for compiling generic programs, everyone is more or less happy with them.
One day you read on a HN that SpaceX satellites have to ship with 4 redundant CPUs. Every CPU costs $100K and consume a lot of power, but they are needed to correct errors introduced by cosmic rays.
You decide to create a compiler that creates programs with error correction automatically embedded. Using your compiler, SpaceX saves $100K and can have 20% smaller solar panels because they can ship it with 3 CPUs instead of 4.
1. How much $ do you think you can sell a single license?
2. How many customers will you actually need to be profitable?
3. How much of your time do you need to create a minimal prototype that you can DM to that SpaceX employee and get him/her in a valuable mutual-help feedback loop?
4. How much better can this compiler become in 1 year of refinement?
Chances are you can make as much you currently make in a year with your first 2 customers.