Agentic toolkit for building and optimizing high-performance inference engines tailored to your GPU and desired LLM.