Neural Networks: Zero to Hero
From scalar autograd and backpropagation in pure Python up to modern Transformer language models built line-by-line from scratch.
Founder of Eureka Labs, Former Director of AI at Tesla
Complete university lecture series taught by Andrej Karpathy.
From scalar autograd and backpropagation in pure Python up to modern Transformer language models built line-by-line from scratch.
Self-contained public lectures, keynote addresses, and research seminars.
Andrej Karpathy introduces Stanford's CS231n course, providing an overview of convolutional neural networks and their historical context in visual recognition.
An foundational exposition on the backpropagation algorithm and the initial principles of neural network architecture, delivered as part of Stanford's CS231n course.
End-to-end engineering dissection of tokenization, distributed Transformer pretraining, supervised fine-tuning, reinforcement learning on chain-of-thought, and inference failure modes.
Andrej Karpathy delivers a keynote at CVPR 2020, detailing the challenges and strategies for achieving scalability in autonomous driving systems.



Linear geometry, convex duality, statistical learning, and deep optimization
MIT · Stanford University · Caltech +1
Search or browse every recorded session across courses and talks.