Show HN: Tiny-vLLM – high performance LLM inference engine in C++ and CUDA

Tiny-vLLM is an educational high-performance LLM inference engine written in C++ and CUDA that implements core optimization techniques like KV caching, paged attention, and continuous batching. The project combines both full source code and a comprehensive course designed to teach developers how to build an efficient inference server from first principles. It supports loading real models (such as Llama 3.2) and performing complete LLM forward passes with CUDA-optimized kernels.