fancyboi999/ai-engineering-from-scratch-zh

This project is a simplified-Chinese version of the open-source course "AI Engineering from Scratch". It's a complete, structured curriculum for learning how AI actually works — not just how to call APIs. The idea: you write the math and code for each algorithm yourself, then use the production libraries (like PyTorch) to do the same thing, so you truly understand what's happening under the hood.

What it includes

• 542 lessons organized into 20 phases, starting from linear algebra and ending with multi-agent systems and production infrastructure.
• Written in Chinese, with English technical terms kept when needed (e.g., agent, token, transformer).
• Covers four programming languages: Python, TypeScript, Rust, and Julia.
• Each lesson lives in its own folder with:
  • runnable code (implemented from scratch)
  • a Chinese document explaining the concepts
  • quizzes to check understanding
  • outputs — a reusable artifact you can take away (a prompt, a skill, an agent, a MCP server).

The whole course follows a “build it, then use it” rhythm: first you code the algorithm using raw math, then you repeat it with PyTorch or scikit-learn, so the framework never feels like magic.

How to use it

There are three ways to start:

1. Read online – free at aieng-zh.cn with full text, interactive diagrams, and progress tracking.
2. Clone and run locally – install the repository and follow the scripts.
3. Let an AI tutor guide you – the project includes built-in “skills” you can add to coding agents like Claude Code or Codex with one command. The tutor assesses your level, creates a personalized learning plan, then teaches you lesson by lesson with quizzes and experiments.

Extras

• 78 hands-on projects (with 322 stages) give you real-world exercises like building a retrieval-evaluation lab or a tool-call firewall.
• Certification prep for Claude (free, community-made) and for the MCP Associate exam.
• The project is a derivative work of the original English course by Rohit Ghumare, used with permission under the MIT license.

In short: it’s a deep, hands-on, all-in-Chinese path to becoming an AI engineer by writing nearly everything yourself, with optional AI tutoring to keep you on track.