Yeah that's also what I'm doing. Also (re-)discovered conda lately. The good thing about is that envs are global, so they can be shared between projects (if you so wish). But maybe more geared towards ML?