Unfortunately Google hasn't hired a million programmers yet so perhaps their ML algorithms aren't good yet at knowing what actually makes people good. In fact,

Unfortunately Google hasn't hired a million programmers yet so perhaps their ML algorithms aren't good yet at knowing what actually makes people good. In fact, I think it is actually rather fool-hardy when I hear people talking about analyzing resumes to predict candidates, or other such nonsense. Just interview some people. Don't ask crazy algorithm questions either, watch the person do what they are suppose to do. It worked great while I was at Pivotal Labs. We would sit down with engineers and pair program with them for a while. It became rapidly clear whether or not they were good, mostly because we were evaluating them on the exact criteria for which we were searching.