Protein structure prediction (admittedly, humans are surprisingly good at this, but AlphaFold is better)
Protein folding problems are cool and all, but as a computational biologist, I do think people underrate simpler ML models on a variety of problems. Researchers have to make a lot of decisions - such as which hits to follow up on in a high throughput screen. The data to inform these kinds of tasks can be expressed as tabular data and decision trees can perform quite well!
I don't have much successful experience but I am (slowly) working my way through Musicians Way https://www.musiciansway.com/practice/ which has some good ideas and worksheet templates for deliberate practice. I have frequently thought that integrating SRS into my music practice would help recall (trying to expand my memorized repertoire as well) but I haven't implemented anything there yet. Would love to hear other ideas.