`ai-engineering-from-scratch`: 435 lessons for end-to-end AI engineering

A full-stack curriculum links math, LLMs, agents, MCP, and production. Too long for quick wins, useful as a structured roadmap.

• —435 lessons, 20 phases, about 320 hours; scope runs from linear algebra to multi-agent systems.
• —Each lesson ships a reusable artifact: prompt, skill, agent, or MCP server; output is more practical than passive reading.
• —Languages span Python, TypeScript, Rust, and Julia; strong breadth, but heavy time cost for product-focused builders.
• —MIT-licensed and laptop-oriented. Better as a reference path than a sprint plan.