You don’t have much time left so you have to act fast.
Find hard problems in areas that interest you to work on, and don’t spend even a second thinking of whether it’s useful for your future career. You can’t predict the future, so you’ll probably be wrong anyway and the time will be wasted.
The point is to get good at solving hard problems in general. I’m on my third Spring project in a row now, and every project has so many NIH idiosyncrasies that no skills transfer at all. The only actually useful skill as a pro dev is to pick up weird, unexpected and even misguided code fast and be productive.