AnthropicPretraining Researcher
Until Present
Building a team using Claude to accelerate pretraining research.
9 unread notifications
Andrej Karpathy researches AI at Anthropic after helping found OpenAI and leading Tesla’s computer-vision team. He also founded Eureka Labs and co-created Stanford’s CS231n course, with educational work that explains neural networks through code.
Andrej Karpathy researches AI at Anthropic after helping found OpenAI and leading Tesla’s computer-vision team. He also founded Eureka Labs and co-created Stanford’s CS231n course, with educational work that explains neural networks through code.
Career
OpenAIResearch Lead, Midtraining and Synthetic Data
2023 — 2024
Returned to OpenAI to build a team working on midtraining and synthetic data generation.
TeslaDirector of Artificial Intelligence
2017 — 2022
Led Tesla's computer-vision team for Autopilot and briefly worked on Tesla Optimus.
Stanford UniversityCS231n Instructor
Jan 2016 — Mar 2016
Co-created CS231n with Fei-Fei Li and was its primary instructor in 2015 and 2016. Dates reflect the Winter 2016 course offering.
Google DeepMindResearch Intern
2015 — 2015
GoogleResearch Intern, Google Research
2013 — 2013
University of British ColumbiaTeaching Assistant, Graduate Probabilistic Machine Learning
2011 — 2011
Teaching assistant for CPSC540; previously taught tutorial sections for UBC’s discrete mathematics course.
Cosigns Received
Featured Media
youtube.com
Deep Dive into LLMs like ChatGPT
youtube.com
Neural Networks: Zero to Hero
github.com
nanochat: training an LLM from start to finish
An educational, open-source stack covering tokenization, training, evaluation and chat inference for language models.
karpathy.bearblog.dev
2025 LLM Year in Review
Karpathy surveys changes in reinforcement learning, LLM applications and coding agents during 2025.