Build a GPT Model From Scratch

​Build a GPT Model From Scratch

​This workshop is an end-to-end walkthrough of how GPT-style language models actually work, without hiding behind libraries or abstractions.

​We’ll start from first principles and progressively build a transformer-based language model from scratch: tokenization, embeddings, self-attention, positional encodings, and the full training loop. You’ll learn not just what each component does, but why it exists and how design choices impact performance, scaling, and behavior.

​This workshop will be taught by Saad Jamal, Co-Founder & CEO of Claybird (YC F25), previously MTS at Essential AI where he worked under the main author of the "Attention is All You Need" paper - Ashish Vaswani. Prior to that, Saad worked on ML systems for autonomous driving at Nuro and Tesla.