The "ai-engineering-from-scratch" repository by rohitg00 is a comprehensive, open-source curriculum designed to bridge the professional gap in AI tool proficiency through hands-on technical implementation. Spanning 20 phases and 523 lessons, the curriculum covers a vast technical spectrum from linear algebra and machine learning foundations to advanced topics like Transformer architectures, Diffusion models, Reinforcement Learning, and autonomous agent swarms. Unlike theoretical courses, this program emphasizes building from first principles; students implement core algorithms—such as backpropagation, tokenizers, and attention mechanisms—manually before utilizing production frameworks like PyTorch or JAX. The curriculum supports Python, TypeScript, Rust, and Julia, and utilizes a "Build It / Use It / Ship It" methodology where every lesson produces a reusable artifact, such as a prompt, an agent, or a Model Context Protocol (MCP) server. Learners can interact with the material via a terminal-based tutor in compatible hosts like Claude Code or Codex, which provides interactive instruction, quizzes, and progress tracking in files like LEARNING.md. The project also serves as a preparation resource for the Claude Certification Academy, offering structured paths for Associate, Developer, and Architect levels.