My supervisor wants to do ML on materials science and since a lot of it has been said and done (and is horrid...), he wanted to add some secret buzzword sauce. Unfortunately he has no idea about ML and what he wants to achieve, he isn't even able to tell me a question I should answer plainly in text. It always goes back to 20min "talks", where he assures me of my good work and asks some questions, a clueless reviewer in peer review would come up (i.e. produce a graphic of X, where X is not relevant to the problem, but related and might be nice to check out, to be 110% sure...).
Disclaimer: if one needs a PhD-dropout physicist with a bit of python and .NET (and a solid understanding of Linux as run in HPC) and an idea about most of the other classic ecosystems, tell me! (no adtech, weapons or cars)