I think it sounds like an excellent “deep dive” project, but don’t get your hopes up that a million people will use it. Ultimately I believe this will be a test of your statistics and math skills more than anything,
I think you could probably reasonably complete the application itself in 1-3 months, but I have no idea how long it would take to complete the matching/recommendation algorithm.
Honestly I just open up a new project and start fucking around until I figure out syntax, language features, and how files relate to each other