# Andrej Karpathy

Research and product · Pretraining research at Anthropic · joined 2026

An OpenAI founding member and Tesla's former head of AI. He joined Anthropic in May 2026 to get back to research.

Full profile: https://www.autumn.ai/resources/anthropic/people/andrej-karpathy

> “I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D.” [1](https://www.cnbc.com/2026/05/19/anthropic-hires-openai-cofounder-andrej-karpathy-former-tesla-ai-lead.html)

**At Anthropic:**
- May 2026 – present: Pretraining research team lead
Sources: [1](https://en.wikipedia.org/wiki/Andrej_Karpathy)

**Career before Anthropic:**
- 2024–2026: Eureka Labs, Founder
- 2023–2024: OpenAI, Returned for a second stint
- 2017–2022: Tesla, Director, then Senior Director of AI
- 2015–2017: OpenAI, Research scientist, founding member
Sources: [1](https://en.wikipedia.org/wiki/Andrej_Karpathy)

**Education:**
- BSc, Computer Science and Physics, University of Toronto, 2009
- MSc, Computer Science, University of British Columbia, 2011
- PhD, Computer Science, Stanford University, 2016
Sources: [1](https://en.wikipedia.org/wiki/Andrej_Karpathy)

**Lesser-known facts:**
- Became known on YouTube first, for Rubik's cube tutorials on his channel badmephisto. [1](https://en.wikipedia.org/wiki/Andrej_Karpathy)
- Designed and taught CS 231n, Stanford's first deep learning class, which grew from 150 students in 2015 to 750 in 2017. [1](https://karpathy.ai/)
- Did his PhD under Fei-Fei Li at the Stanford Vision Lab. [1](https://karpathy.ai/)
- First got into deep learning at the University of Toronto, attending Geoff Hinton's class and reading groups. [1](https://karpathy.ai/)
- At Tesla, his team handled in-house data labeling, neural network training and deployment on Tesla's custom inference chip. [1](https://karpathy.ai/)
- The model cited in his original February 2025 vibe coding tweet was Anthropic's. [1](https://fortune.com/2026/05/19/who-is-andrej-karpathy-vibe-coding-anthropic-openai-rubiks-cube/)
- In March 2026, let an AI agent tune training code unsupervised for two days. Its 700 experiments cut a larger model's training time by 11%. [1](https://fortune.com/2026/05/19/who-is-andrej-karpathy-vibe-coding-anthropic-openai-rubiks-cube/)

**Interviews and talks:**
- [Sequoia Ascent fireside chat (2026)](https://x.com/karpathy/status/2049903821095354523)

**Known for:**
- CS 231n: Convolutional Neural Networks for Visual Recognition
- Connecting Images and Natural Language (2016)
- Eureka Labs / LLM101n
- "Zero to Hero" series
- Coined "vibe coding" (2025)
Sources: [1](https://en.wikipedia.org/wiki/Andrej_Karpathy)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people