Hands-on course: build your first small language model in 30 days. Four weeks cover PyTorch, tokenizers, attention and KV cache, then building and training an SLM from scratch. Labs included; separate GPU needed to train/run a full LLM. $49.