Currently engaged in optimizing llm models on edge devices