[tedmiston]

You can't do much machine learning without users and data, or at least real world datasets. I would focus on picking off the small problems you can solve today (or this week) for now.