Free open-source AI engineering course makes you build backprop, attention and agent loops from raw math
ghumare64 · x · 2026-11-09
"AI Engineering from Scratch" is a free, MIT-licensed open curriculum by Rohit Ghumare with tens of thousands of GitHub stars.
- Philosophy: build everything from raw math before touching frameworks — backprop, tokenizers, attention, and the agent loop are all hand-implemented before PyTorch appears; the path runs from linear algebra to autonomous swarms
- Every lesson is implemented in four languages (Python, TypeScript, Rust, Julia) with a fixed loop: read the problem, derive the math, write code, run tests, keep the artifact
- Learn in your terminal via npx skills add rohitg00/ai-engineering-from-scratch — works as a SKILL.md with Claude, Cursor, Codex and other agents, which act as a tutor with placement quizzes and personalized paths
More from coding & agent
- Personal agents can't handle humans' comfortable 'logistical chaos' — builderjaydub · 2026-11-09
- kitlangton proposes content-addressed compilation units so coding agents can share work across machines — kitlangton · 2026-11-09
- Fixed memory rules hurt strong agents: "record only failures" costs Opus 5 gains — curious_vii · 2026-11-09
- Building a language for agents: Blood makes regex and SQL typos compile-time errors — kitlangton · 2026-11-09
- macOS 27.3 lets Mac VMs use host GPU and Neural Engine, up to 7× faster Core AI — SumitGup · 2026-11-09
- McKinsey: Top 20-25% of Companies See 2x AI Coding Gains — Here's What They Do Differently — AI Engineer · 2026-11-09