Mine isn't from academia specifically. It was along the same vein though. I wanted to figure out how different components worked and ended up making my own full

Mine isn't from academia specifically. It was along the same vein though. I wanted to figure out how different components worked and ended up making my own full stack NLP platform exploring every topic (question answering, named entity recognition, summarization,...). Within about nine months or so, I got it done enough (fully implemented, maybe not perfect like a full team of engineers could do) to build things on top of it. However, it's accurate enough relative to competing or similar products. I decided to think of ways of making money on it, and just built something on top of it, now I have customers and all that fun stuff. A research project and a startup (software obviously) both have code parts to them. It really comes down to the problems you're attempting to solve. In my case, I love the position I'm in now since I have the ability to make a lot of different kinds of products by just adding a web app in front of a rest API. As long as it solves a particular need, customers don't care how or even when you built it as long as it does what's advertised. Research projects are how startups can start though. That's why I think they say get out there and just build something interesting to you (in this case a research project) and see where it takes you.