CookLLM

CookLLM helps developers deeply understand and build large language models from scratch. Instead of using black-box APIs, you write every component yourself — BPE tokenizer, transformer architecture, CUDA/Triton GPU kernels, Flash Attention, pretraining pipeline, and post-training (SFT/RLHF) — guided by interactive visualizations that make complex concepts intuitive.

Software engineers and ML practitioners who want to go beyond surface-level LLM usage. Ideal for developers with Python and basic PyTorch experience who want to understand how LLMs actually work at the systems level whether transitioning from other AI fields, preparing for ML engineering roles, or simply driven by curiosity to build things from first principles.