This is my problem with all the webcam startups. So what if you can mask some of the problems of small sensors and lenses using machine learning that adds a whole new set of problems? You could have done that without even making hardware at all. We have plenty of crappy hardware out there already, and if yours is only a minor improvement with the "magic" in software then it mostly amounts to a hardware dongle to enforce your software license. No thanks!
If you're going to bother making hardware, you should make good hardware. That means a big sensor and a big lens. Start there, and sure go crazy with the machine learning afterward, you'll get much better results with less effort when you start with better input! And you'll have no competition because there's literally nobody else out there putting decent lenses on webcams.