Stanford's CS336 Course Made Public

stanfordnlp · x · 2026-07-15

This is a reshare of Stanford's CS336: Language Modeling from Scratch course, which focuses on building large language models entirely from the ground up.

The curriculum covers data collection and cleaning, Transformers and MoE, GPU and kernel optimization, distributed training, inference, and alignment/RL. It emphasizes hands-on implementation of the entire training pipeline. The original poster describes it as a "pure signal, low noise" course, perfect for those looking to systematically solidify their LLM foundations.

Original post →

More from Companies & People

Companies & People channel →