LLM-Infra-Lab provides small, readable, reproducible demos of real infra primitives (vLLM-style KV cache mock, batching simulator, minimal router/workers, JAX pmap model, etc.), all runnable on CPU/Colab. It’s meant as a learning & experimentation lab for anyone who wants to see how LLM systems actually work under the hood.

Happy to answer questions or add modules people request.