⬢github C++ · 227 ★ +210 since we first saw it · pushed 18 d ago · NOASSERTION
samir-sarwar/samir-vllm
small LLM inference engine from scratch, w/ 6+ hours of youtube tutorials & 85+ pages of handwritten notes included
An educational project that teaches LLM inference from first principles by building a small C++/CUDA inference engine for Llama-style models. It walks through loading SafeTensor weights, embeddings, RMSNorm, RoPE, attention, SwiGLU, KV caching, and batching, accompanied by 6+ hours of YouTube tutorials and 85+ pages of handwritten PDF notes.
Why now: It's being discussed on Hacker News for its unusually complete teaching materials — hours of video plus extensive hand-written notes on LLM internals.
Who it is for: Developers and students who want to understand how transformer inference works at the GPU/kernel level, beyond just calling APIs.
Stars over our 1042 snapshots: 17 to 227, since 12 d ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/samir-sarwar/samir-vllm