Andrej Karpathy
Pre-training team, Anthropic; OpenAI co-founder
Education and Stanford
Andrej Karpathy earned a BSc at the University of Toronto from 2005 to 2009, with a double major in computer science and physics[1]. He writes that he first got into deep learning there by attending Geoff Hinton's class and reading groups[1]. From 2009 to 2011 he took an MSc at the University of British Columbia, working with Michiel van de Panne on learning controllers for physically simulated figures[1].
His PhD at Stanford, from 2011 to 2015, focused on convolutional and recurrent neural networks applied to computer vision and natural language processing, with Fei-Fei Li as his adviser[1]. During the PhD he interned at Google Brain in 2011, Google Research in 2013 and DeepMind in 2015[1]. With Fei-Fei Li he wrote Deep Visual-Semantic Alignments for Generating Image Descriptions, posted to arXiv in December 2014[2].
Karpathy designed and was the primary instructor for CS231n, Stanford's first deep learning class, on convolutional neural networks for visual recognition[1]. Enrolment grew from 150 students in 2015 to 330 in 2016 and 750 in 2017[1].
OpenAI and Tesla
Karpathy was a research scientist and founding member at OpenAI from 2015 to 2017[1]. He was listed among OpenAI's founding members when the non-profit was announced in December 2015[3].
From 2017 to 2022 he was director of AI at Tesla, where he led the computer vision team for Autopilot and, briefly, Tesla Optimus[1]. His team handled in-house data labelling, neural network training and deployment on Tesla's custom inference chip[1]. TechCrunch reported in July 2022 that he was leaving Tesla[4].
He returned to OpenAI from 2023 to 2024, where he built a team working on midtraining and synthetic data generation[1].
Eureka Labs and teaching
On 16 July 2024 Karpathy announced Eureka Labs, which described itself as building a new kind of school that is AI native[5]. Its model pairs teacher-designed course materials with an AI teaching assistant, and its first planned product was LLM101n, an undergraduate-level course in which students train their own AI[5].
Since 2024 he has made educational videos on his YouTube channel in two tracks: a technical track built around the Neural Networks: Zero to Hero playlist, and a general audience track that includes Deep Dive into LLMs like ChatGPT[1]. He has also written widely read blog posts, including The Unreasonable Effectiveness of Recurrent Neural Networks in May 2015 and Software 2.0 in November 2017[1].
He publishes small open-source code bases for teaching and experiments. nanoGPT is described as the simplest, fastest repository for training and fine-tuning medium-sized GPTs[6]. nanochat is a minimal harness for training LLMs on a single GPU node, covering tokenisation, pretraining, fine-tuning, evaluation and inference[7]. Its README says it can train a model with GPT-2 capability for about $48 in roughly two hours on an 8-GPU H100 node[7].
Joining Anthropic
On 19 May 2026 Karpathy announced that he had joined Anthropic, where he works on pre-training under team lead Nick Joseph[8]. An Anthropic spokesperson told TechCrunch that he would start a team focused on using Claude to accelerate pre-training research[8]. In his announcement he said he remained passionate about education and planned to resume that work in time[8]. TechCrunch reported that it was not clear whether he would continue with Eureka Labs[8].
Known for
- Founding member of OpenAI
- Director of AI at Tesla, leading Autopilot computer vision
- Creating Stanford's CS231n course
- Educational videos and code such as nanoGPT and nanochat
- Founding Eureka Labs
Sources
- Andrej Karpathy's website: career, teaching and writing — checked 6 Nov 2026
- arXiv: Deep Visual-Semantic Alignments for Generating Image Descriptions, December 2014 — checked 6 Nov 2026
- TechCrunch: Nonprofit OpenAI launches, December 2015 — checked 6 Nov 2026
- TechCrunch: Tesla loses top AI executive who led Autopilot vision team, July 2022 — checked 6 Nov 2026
- Eureka Labs: Introducing Eureka Labs, July 2024 — checked 6 Nov 2026
- GitHub: karpathy/nanoGPT — checked 6 Nov 2026
- GitHub: karpathy/nanochat — checked 6 Nov 2026
- TechCrunch: OpenAI co-founder Andrej Karpathy joins Anthropic's pre-training team, May 2026 — checked 6 Nov 2026