Experimental design and observational causal inference would be excellent skills to have. Especially if you’re working with people who are asking you “why” questions, ML is helpful but isn’t going to cut it alone.