Ive been studying ML for a while and the best way that seems to work for me is by working on smallscale experiments every 2 weeks. Havent found any hackathons yet, otherwise that’d be a good way as well.