Researchers from Team: William Hu, Drew Wadsworth, Sean Siddens, Stanley Winata, Daniel Fu, Ryan Swann, Muhammad Osama, Christopher Ré, Simran Arora, developed HipKittens to optimize AI performance on AMD GPUs. HipKittens uses optimized memory access, AMD-centric wave scheduling, and chiplet-aware grid scheduling to achieve competitive performance on AMD CDNA3 and CDNA4 GPUs.