Makora (formerly Mako) builds inference endpoints that use AI agents to automatically optimize GPU performance across every layer of the stack — from kernels and inference engines to orchestration — so teams can run models faster and cheaper without manual tuning.