Make frontier model inference faster end-to-end. You will work across GPU kernels, compilers, runtimes, batching, scheduling, and serving to turn systems research into production performance.
