Jay Alammar
Also known as The Illustrated Transformer
Jay Alammar teaches machine learning through visual explanations and worked examples. His 2016 interactive guide introduced neural networks through small models and visible changes. The Illustrated Transformer followed in 2018, using diagrams to explain attention and model structure. He then extended the approach to GPT-2 and Stable Diffusion. With Maarten Grootendorst, he co-authored Hands-On Large Language Models, published in September 2024. The book’s public code covers embeddings, classification, prompting, retrieval, and fine-tuning. His public relevance comes from a sustained body of educational material that helps readers connect model components with their behaviour. The illustrated articles and shared book repository give learners diagrams, explanations, and code they can inspect together.
Timeline
-
December 2016
Making neural networks interactive
Alammar published a visual and interactive introduction to neural networks in December 2016. It used small examples to connect predictions, weights, and training. The article shows an early teaching method: make the machinery visible and let readers inspect how changes affect a simple result.
-
June 2018
Illustrating the Transformer
The Illustrated Transformer broke the architecture into diagrams of attention and its surrounding components. Alammar connected the parts to the flow of information through the model. The article teaches ideas from the original research paper.
-
2019
Explaining a generative language model
The Illustrated GPT-2 followed the decoder-only model through its architecture and text generation. It built on the visual language of the Transformer article. Readers could connect tokens, attention, and the next-word process without treating the system’s generated language as proof of human-like understanding.
-
2022
Explaining text-to-image generation
The Illustrated Stable Diffusion applied the same visual method to image generation. It described a model generating an image from text and explained the components involved. The article’s update notes record later changes to the diagrams, making correction and refinement part of the teaching record.
-
September 2024
A shared book and code collection
Alammar and Maarten Grootendorst published Hands-On Large Language Models in September 2024. The book covers understanding, using, and training language models. Its official repository provides examples for topics such as embeddings, classification, prompting, retrieval, and fine-tuning.
-
Current public material
A library of visual explanations
The articles and book code remain public references for developers and learners. Their sequence connects foundational network ideas with modern language and image models. Alammar’s role is that of an explainer who helps people inspect technical structure and then try the ideas through worked examples.
Sources and credits
Photo: Jay Alammar / personal website. License: Not stated by source. Source: jalammar.github.io. Cropped to a square and resized. Ask us to remove or replace this photo.