AI Engineering from Scratch is a comprehensive, open-source curriculum designed to teach artificial intelligence development from foundational mathematics through production-ready systems. Created by rohitg00, the project is built around the philosophy that practitioners should understand how AI actually works by building algorithms from scratch rather than simply calling APIs. The curriculum spans 503 lessons organized across 20 phases, covering approximately 320 hours of material in Python, TypeScript, Rust, and Julia.
The repository addresses a documented gap in AI education: while 84 percent of students use AI tools, only 18 percent feel professionally prepared to do so. The curriculum closes this gap by teaching linear algebra at the foundation and progressing through to autonomous swarms at the advanced end. Each lesson follows a consistent structure where learners first derive the mathematics, implement the algorithm from raw code, run tests, and retain a reusable artifact. This approach ensures that by the time production libraries like PyTorch are introduced, learners already understand the underlying mechanics.
The curriculum is organized into 20 stacked phases, beginning with setup and tooling in Phase 0, moving through math foundations, and progressing to advanced topics like agents and production systems. The structure is intentionally linear, with each phase building on previous knowledge. Learners can skip ahead if they already understand lower layers, but the curriculum warns against skipping foundational material and then encountering problems at higher levels.
A distinctive feature of the project is that every lesson produces a reusable artifact. Rather than ending with congratulations on learning a concept, each lesson ships something practical: prompts for AI assistants, skills that integrate with Claude and other agents, autonomous agents built from scratch, or MCP servers. By the end of the curriculum, learners accumulate a portfolio of 503 artifacts they genuinely understand because they built them themselves.
The project includes built-in agent skills for popular AI tools. The find-your-level skill provides a ten-question placement quiz that maps learner knowledge to an appropriate starting phase and generates personalized learning paths with hour estimates. The check-understanding skill offers per-phase quizzes with feedback and specific lesson recommendations for review.
According to GitGenius activity tracking, the repository shows strong maintenance patterns with a median issue and pull request response latency of 0.0 hours and a mean latency of 15.4 hours across 41 tracked items. The primary maintainer, rohitg00, has logged 26 events, with secondary contributors abhinav-m22 and alexisvalentino each contributing 4 events. Bug reports represent the most active issue label with 20 tracked items, followed by help wanted with 2 items and good first issue with 1 item. The repository connects to related projects including rohitg00's Agent Memory project, which provides persistent memory for AI agents and chat assistants.
The project has achieved significant reach, with 150,639 readers and 241,669 page views in a recent 30-day period. The curriculum is free, open source under the MIT license, and designed to run on personal laptops without requiring expensive hardware or cloud infrastructure. The accompanying website at aiengineeringfromscratch.com provides access to all lessons without requiring repository cloning, though learners can also clone and run locally or use agent-based placement testing to find their optimal starting point.
