Andrej Karpathy: AI Engineering & Deep Learning

Ravenza Editorial

352 installs · 29 sources · Updated 8/18/2026

Andrej Karpathy is an AI researcher and educator known for making neural networks and language models understandable from first principles. Use this Skill for grounded explanations of deep learning, LLMs, tokenization, training, evaluation, and agentic software.

deep-learningneural-networkslanguage-modelstokenizationtransformersllm-engineeringcoding-agentsevaluation

Example questions

→Explain backpropagation through a tiny autograd graph.

→Why are my activations saturated and gradients unstable?

→How does byte-pair encoding build a tokenizer?

→Walk me through self-attention from first principles.

→Help me reproduce a small GPT before modifying it.

→What should I verify when using an LLM for coding?

Related Skills

Sources

29 web pages · 10 verified

Software 2.0karpathy.medium.com llm-wikigist.github.com Building makemore: language modelingyoutube.com Building makemore Part 2: MLPyoutube.com Building makemore Part 3: activations, gradients, and BatchNormyoutube.com Building makemore Part 4: becoming a backprop ninjayoutube.com

Update history

Skill details refreshed · 8/18/2026

The title, description, or cover was improved.

Initial release · 8/17/2026

First published on Ravenza.