If it's for practice or portfolios I tend to suggest people solve problems which have already been solved. Find a SaaS service which already exists and just write your own naive version and open source it. Trello, Hacker News, a doc viewer, etc.

If you're looking for a problem to work on which will allow you to sell your solution later on, that's a different kettle of fish.

For me, the number one issue when learning a programming language is motivation so I chose to pick a problem I personally need solved so I'll feel the need and have the motivation to see it through.