Training LLMs

This self-paced course teaches absolute beginners how large language models are trained — from raw text to a useful, evaluated, and safely deployed assistant — with no advanced math, built around the master recipe of pretraining (language), fine-tuning (skill), and alignment (judgment). It builds understanding step by step through tokens and next-token prediction, the Transformer's attention mechanism, the data pipeline, evaluation, and safety, all anchored by the practical "intervention ladder" — prompt → retrieve → fine-tune → align, cheapest first. It closes hands-on with two quizzes, four no-coding exercises, a capstone to design a one-page training plan, a printable takeaways slide, and a glossary.