FIG_000 · curriculum v1.0 · 2026 open source · MIT

AI Engineering
from Scratch

Every published lesson. Every phase. Every algorithm built from raw math before a single framework gets imported.

Maintained by Rohit Ghumare and contributors. Run on your own machine.

Start the Course Explore Learning Paths Star on GitHub … Follow @rohitg00
Learn in your terminal 
npx skills add rohitg00/ai-engineering-from-scratch
> Use start-learning to begin the course.
Claude Cursor Codex + any SKILL.md agent

Your agent becomes your tutor: placement quiz, personalized path, lessons taught interactively in your terminal.

FIG_001 · FORWARD PASS y = σ(Wx + b) x h₁ h₂ ŷ
Plate 1 of 3. Every layer of this diagram is a lesson you implement by hand.
Read by engineers and students at
Apple Google Meta OpenAI NVIDIA IIT Bombay University of Windsor

“Obsessed with the AI Engineering from Scratch repo.” - AI engineer at Google

How this works

Most AI material teaches in scattered pieces. A paper here, a fine-tuning post there, a flashy agent demo somewhere else. The pieces rarely line up. You ship a chatbot but can't explain its loss curve. You hook a function to an agent but can't say what attention does inside the model that's calling it.

This curriculum is the spine. Every phase, every published lesson, four languages: Python, TypeScript, Rust, Julia. Linear algebra at one end, autonomous swarms at the other. Every algorithm gets built from raw math first. Backprop. Tokenizer. Attention. Agent loop. By the time PyTorch shows up, you already know what it's doing under the hood.

Each lesson runs the same loop: read the problem, derive the math, write the code, run the test, keep the artifact. No five-minute videos, no copy-paste deploys, no hand-holding. Free, open source, and built to run on your own laptop.

Choose the work you want to do

AI engineering is larger than model code. Choose one of four core learning paths, then learn from the same source, labs, tests, and artifacts in the browser or on GitHub.

View Learning Paths Browse career routes
AI Engineering 4 connected domains
AI Engineering connects building and deploying AI applications, software engineering fundamentals, agent-assisted engineering, and product judgment and delivery. Building and Deploying AI Applications Software Engineering Fundamentals Agent-Assisted Engineering Product Judgment and Delivery
Recommended firstNew to AI engineering

Set up a working environment, run the repository, and learn the lesson workflow before choosing a specialization.

Open lesson GitHub source
Core domainBuilding and Deploying AI Applications

Move from prompts, structured outputs, embeddings, and retrieval through evaluation, serving, observability, and safe release.

Start path GitHub path
Core domainSoftware Engineering Fundamentals

Build the repository, environment, interface, debugging, verification, security, release, and operational foundations AI systems depend on.

Start path GitHub path
Core domainAgent-Assisted Engineering

Frame the task, plan from repository evidence, engineer the loop and harness, isolate delegation, verify the result, and preserve feedback.

Start path GitHub path
Core domainProduct Judgment and Delivery

Turn observed work into outcomes, assumptions, testable slices, executable specifications, measurement plans, staged releases, and owned feedback.

Start path GitHub path
Focused pathModel Context Protocol (MCP)

Build, secure, verify, and operate stateless MCP systems from wire envelopes through release gates.

Start path GitHub source
Focused pathAgent Skills

Build, invoke, route, secure, evaluate, package, and verify portable skills in real agent hosts.

Start path GitHub source
Practice by evidenceCertification preparation

Choose a certification route, complete practical labs, keep learner-owned artifacts, and use original assessments.

Explore paths GitHub tutor
Current Progress
Finished Lessons 0 / 0
Phases 0 / 0
Languages 4
Glossary Terms ···
Curriculum · 20 phases · 523 lessons
Tap a phase to expand its lessons. Each one ships when its math, code, and test are all written.
Complete In progress Planned
The book edition · six volumes
The course, compiled. EPUB and PDF built from the same lessons and attached to every GitHub release. The site stays the living edition. Every chapter links back here for the animated figures, quizzes, and code.
Links resolve to the newest GitHub release · rebuilt by CI from the lessons on every release · how it's made
Independent certification preparation

Prepare by building the real systems

Four Claude certification paths taught the same way as the course: step by step, with interactive labs, practical artifacts, and an AI tutor that works from the GitHub repo.

4 tracks 33 certification lessons 295 original practice questions

Not affiliated with, endorsed by, sponsored by, or authorized by Anthropic. This curriculum does not issue credentials or guarantee a passing result.

Explore certifications Learn with the GitHub tutor
Colophon

The entire curriculum is on GitHub. Clone it, fork it, learn at your own pace. No paywall, no signup. Every lesson has runnable code in Python, TypeScript, Rust, or Julia, depending on what fits the concept best.

git clone https://github.com/rohitg00/ai-engineering-from-scratch.git