I think about efficient compute. I spend time on hardware, systems, ml, and their intersection(s). Currently: efficient inference @ Sail Research hardware efficiency, co design, world-action research @ Google model architecture, reasoning, kernels @ Dao AI Lab