CS224n: Natural Language Processing with Deep Learning

Christopher Manning's course builds up from word vectors and RNNs to attention and the Transformer architecture behind every current LLM, and the assignments make you implement the ideas rather than just hear them. It is the foundations layer under the applied 'use the OpenAI API' courses, which take the models as given.

It is advanced and maths-heavy, not a first ML course, and there is no certificate from the free materials. If you are new to ML, start with a foundations course; if you only want to use language models rather than understand them, an applied course fits better. The public recordings trail the newest models by a little, but the core ideas still hold (as of 2026).

CS224n is Stanford's NLP with deep learning course, a widely used academic resource for understanding how modern language models work. Taught by Christopher Manning, a leading NLP researcher, it covers word vectors (Word2Vec, GloVe), recurrent neural networks, sequence-to-sequence models, attention mechanisms, the Transformer architecture that underpins every modern LLM, BERT, GPT, and the current landscape of large language models.