AI Engineering from Scratch hits 299 lessons

This is a serious curriculum, not a lightweight tutorial series. The strongest idea is the repeatable workflow: concept, implementation, then artifact, which turns learning into a portfolio instead of passive consumption.

• –The scope is unusually broad for one repo: math, classical ML, deep learning, LLMs, agents, and deployment infrastructure all sit in one path
• –The from-scratch-first structure should help people understand why the abstractions work before they reach PyTorch, sklearn, or inference tooling
• –The built-in Claude Code skills make it more AI-native than most courses, because the repo itself participates in assessment and placement
• –The course is especially useful for engineers who already use AI tools but want to move from prompt consumption to systems they can build and reuse
• –The risk is ambition: 299 lessons can overwhelm learners unless the phase selection and onboarding stay sharply curated