I agree. I did molecular dynamics simulations for my PhD. The last year I spent some time exploring feature representations that capture atomic environments u

I agree. I did molecular dynamics simulations for my PhD. The last year I spent some time exploring feature representations that capture atomic environments up to their symmetry invariants (rotations, reflections, permutations of identical atoms, etc.) The use of machine learning for MD and quantum chemistry has exploded in the last year and a half. I'm a bit sad that I just finished my degree right as this kind of work is heating up.