Jay Alammar is Director and Engineering Fellow at Cohere, where he works on code agents and the harder question of how to tell whether they’re any good: agentic evaluation and observability, making a model’s behavior legible enough to inspect, measure and improve. That work shapes Cohere’s models, including North Mini Code. He joined Cohere in 2021 as a Partner Machine Learning Engineer, helping developers solve business problems with language AI. Jay is best known for The Illustrated Transformer and the visual explanations on his blog, which have helped tens of thousands of people wrap their heads around machine learning, been translated into many languages and been referenced in courses at Stanford, MIT and other universities. With Maarten Grootendorst, he co-authored two O’Reilly books, Hands-On Large Language Models and An Illustrated Guide to AI Agents, and the two created the free course How Transformer LLMs Work with Andrew Ng and DeepLearning.AI. He is a member of The RAG Pack, a group of AI educators teaching how AI works and where it fails, and an advisor and venture partner at Outliers Venture Capital.