LLM inference, fine-tuning open-weight models, tool calling and AI agents, long context, RAG over enterprise documents