>going from "predicting stuff about atoms" to "using super state-of-the-art ML to predict stuff"

At this stage, they typically publish in their field right?

Any advice for phd students in other fields trying to join ML research? Like I took dl classes already, but how do I get into research?