This is a really neat project! I can follow and dupe everything in the writeup until the OP casually mentions:
> To get these numbers I trained two supervised models. I fine-tuned the pipeline and hyperparameters with genetic programming on an AWS GPU-heavy machine.
Can anyone recommend some ways to level up a bit in this domain to feel more comfortable? I'm somewhat familiar and understand the underlying concepts to a certain extent, such as supervised vs nonsupervised, but need hands-on practice, what do you recommend?