Few people have shaped how a generation of engineers learn AI as much as Andrej Karpathy. A researcher who built some of the most visible AI systems in industry, he is equally known for courses, videos and essays that explain neural networks from first principles.

Early life and education

Karpathy was born in 1986 in what was then Czechoslovakia and moved to Toronto as a teenager. He studied computer science and physics at the University of Toronto, completed a master’s degree at the University of British Columbia, and earned his PhD at Stanford University in 2015 under Fei-Fei Li, working on deep learning that connects images and natural language — for example generating descriptions of images.

CS231n

At Stanford, Karpathy co-created and taught CS231n: Convolutional Neural Networks for Visual Recognition with Fei-Fei Li. The course’s lecture videos and notes became one of the most popular free resources for learning deep learning worldwide, at a time when few structured courses existed.

OpenAI and Tesla

Karpathy was a founding member of OpenAI in 2015, working on research including generative models and reinforcement learning.

In 2017 he joined Tesla as director of AI, leading the computer vision team behind Autopilot. He championed a vision-based approach using neural networks trained on data from the fleet, and described in talks how the team built data engines that collect difficult cases from cars to improve models. He left Tesla in 2022.

He returned to OpenAI in 2023 and left again in early 2024.

Eureka Labs and education

In 2024 Karpathy founded Eureka Labs, a company focused on AI-native education, with the idea of combining expert-designed course materials with AI teaching assistants.

His public teaching has had a huge audience:

  • Neural Networks: Zero to Hero, a video series building neural networks and a GPT-style language model from scratch in Python;
  • talks such as an introduction to large language models for general audiences;
  • small, readable open-source projects like micrograd and nanoGPT that show how core systems work.

See our list of YouTube channels to learn AI.

Ideas that spread

  • “Software 2.0” (2017): an essay arguing that neural networks represent a new kind of software, where behaviour is learned from data rather than written by hand.
  • “Vibe coding” (2025): a term he coined for building software by describing what you want to an AI and accepting its code with little manual editing — widely adopted, and widely debated.

Why he matters

Karpathy bridges research, engineering at scale and teaching. His explanations have lowered the barrier to understanding how modern AI works, and his views on agents, education and software are closely followed across the industry.

Learning from Karpathy

A good path through his material:

  1. Watch his introduction to large language models for the big picture.
  2. Follow Neural Networks: Zero to Hero, coding along from micrograd to a small GPT.
  3. Read the nanoGPT code to see how a real training loop is structured.
  4. Revisit the CS231n notes for the foundations of convolutional networks.

His style — start from the simplest working version, then build up — is a useful way to learn any technical subject.

See our free learning path to understand AI and the profile of his PhD advisor’s field-defining dataset in women who shaped AI.

What to watch next

  • Eureka Labs’ first courses and AI teaching assistants.
  • New educational videos on reasoning models and agents.
  • Open-source teaching projects reproducing modern training pipelines.
  • Debates about AI-assisted programming practices.

Key takeaways

  • Karpathy earned his Stanford PhD under Fei-Fei Li and co-created CS231n.
  • He was a founding member of OpenAI and Tesla’s director of AI from 2017 to 2022.
  • In 2024 he founded Eureka Labs, focused on AI-native education.
  • His Zero to Hero videos and projects like nanoGPT teach AI from first principles.
  • He coined “Software 2.0” and “vibe coding”.

Frequently asked questions

What is Andrej Karpathy known for?

Leading AI for Tesla Autopilot, being a founding member of OpenAI, and teaching deep learning through CS231n and popular videos.

What is Eureka Labs?

An AI education company founded by Karpathy in 2024 that aims to pair course materials with AI teaching assistants.

What does “vibe coding” mean?

A style of programming where you describe what you want to an AI and largely accept the code it writes, coined by Karpathy in 2025.