Very cool to see someone come back out of "industry" and to crack a problem. On a side note: the name looked familiar to me, and I realized I had spent a decent amount of time re-implementing his most cited (ML) paper a few years back.