Multi-stage LLM agent pipeline for optimizing Triton kernels on Intel XPU — from analysis to autotuning. - View it on GitHub