Install WSL2, configure GPU passthrough, set up Ollama and llama.cpp with CUDA, and optimize memory for LLM inference. Step-by-step for Windows 11.