AI Engineering from Scratch

A comprehensive open-source curriculum teaching AI engineering through hands-on implementation. The course spans 20 phases covering linear algebra, neural networks, tokenizers, attention mechanisms, agents, and production systems across Python, TypeScript, Rust, and Julia. Each lesson requires deriving mathematics before writing code, with learners building reusable artifacts like prompts, skills, and MCP servers.

Designed for engineers and students who want to understand AI fundamentals rather than copy-paste solutions, the material runs on personal machines and includes an AI tutor agent that personalizes learning paths and teaches interactively in the terminal. The project is maintained by Rohit Ghumare and contributors under an MIT license, with translations available in multiple languages.