← Home

REVIEWED & READY TO INSTALL

AI Engineering from Scratch

rohitg00/ai-engineering-from-scratch

Free, open-source AI engineering curriculum: 523 lessons across 20 phases, building everything from raw math to LLMs, agents and swarms; every lesson ships a prompt, skill, agent or MCP server, with an AI tutor that runs inside your coding agent.

STEP BY STEP

How to install AI Engineering from Scratch

The step-by-step guide is right on this page, written for people who have never installed anything from GitHub.

Updated 9/30/2026
GitHub preview image of AI Engineering from Scratch
GitHub stars
79,698
Language
Python
Trending rank
#10
Added
9/25/2026

How to install AI Engineering from Scratch, step by step

Work from top to bottom. Each grey box is one command: copy the whole line, paste it into the terminal window and press Enter.

Never used a terminal? Read this first (1 minute)

A terminal is a window where you type commands for your computer. You do not need to understand them, just copy and paste them exactly.

  1. Windows: press the Windows key, type "PowerShell", then press Enter.
  2. macOS: press Command + Space, type "Terminal", then press Enter.
  3. Linux: press Ctrl + Alt + T.
  4. Copy a command from the guide, right-click inside the window you just opened to paste it (macOS: Command + V), then press Enter.
  5. Wait for the command to finish (the cursor blinks again on a new line) before running the next one.

Lines starting with # are notes; you do not need to copy them. If red error text appears, check the common errors section at the end of the guide.

1. What is AI Engineering From Scratch?

This is not an app you install. AI Engineering From Scratch is a free, open-source course on how AI systems are actually built, maintained by Rohit Ghumare and the community. It has 523 lessons across 20 phases, starting with the underlying math and classical machine learning, moving through neural networks and large language models, and ending with agents (AI programs that carry out multi-step work on their own) and production operations.

The teaching approach is to write each algorithm by hand from the raw math first, and only then redo it with a standard library. Every lesson leaves you with something reusable, such as a prompt template or a "skill" for a coding assistant.

To be direct about it: most of the course is for people who can already write code. The stated prerequisite is "you can write code (any language; Python helps)". Complete beginners do have an entry point (Phase 0), but the author estimates roughly 306 hours of study from there.

The good news for non-technical readers is that you do not have to install anything to read it. Every lesson is on the course website, the site offers machine-translated versions in several languages, and the whole curriculum is also packaged as six downloadable PDF / EPUB volumes.

The course is MIT licensed, costs nothing, and needs no account.

A good fit if:

  • You want to understand how AI works inside rather than just use a chatbot.
  • You are learning to program, or already work in software and want to move into AI.
  • You like reading, doing exercises and running code yourself.

Not a good fit if:

  • You are looking for an AI app that works as soon as it is installed. There are only lessons here.
  • You want short video lessons. This course is reading plus code, with no five-minute videos.

2. What do you need?

It depends on how you want to study:

Way to studyWhat you needDoes it cost money?
Read on the webA browser, on a phone or a computerNo
Download the PDF / EPUB booksA PDF reader or an e-book appNo
Learn with an AI tutorNode.js and a coding assistant that supports skills (Claude Code, Codex, Cursor and similar)The course is free; the coding assistant may charge separately, so check that tool's own site
Clone the repository and run the codeGit and Python 3.11 or newerNo

A quick word on the underlying tools:

Each book volume is only about 0.3 to 3 MB, so disk space is not a concern. You do not need a graphics card to begin; the author describes the course as built to run on your own laptop.

3. Step-by-step setup

There are four routes, ordered from easiest to hardest. Pick one to start with.

Route 1: Read on the web (nothing to install)

Step 1: Open the course site

Go to https://aiengineeringfromscratch.com in your browser.

You are on track when: you see the course home page with its list of phases.

Step 2: Choose a language and a starting point

Since Edition 2026.10 (released 27 September 2026), both the site interface and the lesson text are available as machine translations in eight languages: Simplified Chinese, Hindi, Spanish, Arabic, French, Portuguese, Turkish and Vietnamese. Code samples stay in their original form.

If you are unsure where to start, open the prerequisites guide: https://aiengineeringfromscratch.com/prereqs.html

You are on track when: you can open a lesson and read its explanation.

Route 2: Download the books for offline reading

Step 1: Open the latest release

Go to https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest

At the time of writing the newest release is Edition 2026.10 (tag v2026.10).

Step 2: Download the volume you want

Scroll to "Assets". There are six volumes, each as a .pdf and an .epub:

VolumeFile nameCovers
1aiefs-vol1-foundationsMath, tooling, classical machine learning
2aiefs-vol2-deep-learningNeural networks, vision, speech
3aiefs-vol3-languageLanguage processing and the Transformer
4aiefs-vol4-llmsLarge language models
5aiefs-vol5-agentsAgents, protocols, multi-agent systems
6aiefs-vol6-productionInfrastructure, safety, capstone projects

Newcomers should begin with volume 1.

You are on track when: the file opens in your PDF reader or e-book app. The books are in English. The certification prep tracks are not included in the books; they live on the website and on GitHub only.

Route 3: Learn with an AI tutor inside your coding assistant

This turns the coding assistant you already use into a tutor: it asks ten placement questions, builds a personal study plan, then teaches one lesson at a time with a quiz.

Step 1: Check that Node.js is installed

node --version
npx --version

You are on track when: each command prints a version number. If the command is not found, install Node.js from https://nodejs.org and reopen your terminal.

Step 2: Install the learning skills

npx skills add rohitg00/ai-engineering-from-scratch

The installer lists the assistants it can configure and asks where to install. Pick the assistant you actually use. This step needs neither a copy of the repository nor Python.

You are on track when: the installer reports that it wrote the skills into your assistant's skills folder, for example .claude/skills/.

Route 4: Clone the repository and run the code

Step 1: Download the repository

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

You are on track when: the download finishes and your prompt is inside the ai-engineering-from-scratch folder.

Step 2: Check your environment

macOS or Linux:

python3 phases/00-setup-and-tooling/01-dev-environment/code/verify.py --route beginner

Windows (if python3 is not recognized):

python phases/00-setup-and-tooling/01-dev-environment/code/verify.py --route beginner

You are on track when: the script lists the tools you need right now and says you are ready to start. If something is missing, it states the reason and the command that fixes it.

4. First use

Reading on the web or in the books: open the first lesson of the phase you chose. The author recommends the same rhythm for every lesson: read it and restate the main idea in your own words, type out the important code yourself, run it, and move on only when you can explain the output.

Using the AI tutor: open your coding assistant and call the onboarding skill. How you call it depends on the assistant:

AssistantStart the course
Claude Codetype /start-learning
Codextype start-learning, or pick it from /skills
Other assistantstype "Use start-learning to begin the course."

The tutor asks ten placement questions and saves your plan to a file named LEARNING.md in your working folder. In later sessions, call the learn skill for the next lesson. When you are stuck on a topic, call course-guide to be pointed at the exact lesson.

Using the cloned repository: run the very first lesson, which needs no extra libraries.

python3 phases/01-math-foundations/01-linear-algebra-intuition/code/vectors.py

The expected result: the program runs to the end and finishes by showing that multiplying a matrix by a vector is the operation inside a neural network layer. The author suggests saving that output as your first piece of learning evidence.

The author's suggested starting points:

Your backgroundStart atEstimated time
New to programming and AIPhase 0about 306 hours
Know Python, new to machine learningPhase 1about 270 hours
Know machine learning, new to deep learningPhase 3about 200 hours
Know deep learning, want LLMs and agentsPhase 10about 100 hours
Senior engineer, only want agent engineeringPhase 14about 60 hours

5. Common problems and fixes

SymptomCauseFix
node or npx gives "command not found" / "not recognized"Node.js is missing, or the terminal has not picked it upInstall Node.js from https://nodejs.org, then close and reopen the terminal
verify.py reports "need Python 3.11+"Your Python is too oldInstall a newer Python. The script suggests uv python install 3.12; you can also download it from https://www.python.org/downloads/
Windows does not recognize python3On Windows, Python usually goes by a different nameReplace python3 with python or py -3 in the commands
A lesson command cannot find its fileYou are in the wrong folderAll lesson paths start at the repository root, the folder containing README.md and phases/. cd there and run it again
The assistant cannot see the skills after installingThe installer wrote to a different folder than the one your assistant readsRun npx skills add ... again, choose the right assistant and scope, then restart the assistant
The repository's own script python3 scripts/install_skills.py <target> (an alternative to npx skills add, run from a cloned copy) stops with exit code 1The destination already contains files with the same namesAdd --dry-run to preview the collisions, or --force to overwrite
A lesson fails because a library is missingThat lesson needs an external packageOpen the code file, find the first-line comment of the form # requires: ... and install the packages it lists

Still stuck? Ask at https://github.com/rohitg00/ai-engineering-from-scratch/issues

6. Uninstalling / updating

Web reading: nothing to remove or update. The site is always the current edition.

PDF / EPUB books: delete the files when you no longer need them. To get a newer edition, download again from the releases page. Each copy carries a dated edition stamp, and older editions remain downloadable.

AI tutor:

  • Update: run the install command again to fetch the newer skills.
npx skills add rohitg00/ai-engineering-from-scratch
  • Remove: delete the skill folders the installer wrote (for example inside .claude/skills/, .cursor/skills/ or .codex/skills/, depending on what you chose). The LEARNING.md file holds your progress; keep it if you plan to continue.

Cloned repository:

To update, run this inside the ai-engineering-from-scratch folder:

git pull

To remove it, delete the whole ai-engineering-from-scratch folder.

7. Frequently asked questions

Does it cost money? No. The course is free and open source under the MIT license. Only the coding assistant used as a tutor (if you take that route) is a separate product that may charge.

I have never programmed. Can I still do this? There is an entry point at Phase 0, but the author states that being able to write code is a prerequisite and estimates about 306 hours for someone completely new. Read a few lessons on the website before committing.

Is it available in other languages? The website offers machine-translated interface and lesson text in eight languages. The PDF / EPUB books and the source code are in English.

Do I need an internet connection? For the website, yes. The books, once downloaded, and the repository, once cloned, can be read offline. The AI tutor needs a connection because the coding assistant has to reach its own service.

Do I get a certificate at the end? The course itself does not issue one. It includes prep material for the Claude certifications and the MCP Associate certification, but states clearly that this is independent study material, not affiliated with the exam bodies, and that it cannot guarantee a pass.

Written by GitHot and checked against the project's official documentation. If a step is wrong or unclear, message GitHot through the channels at the bottom of the page.

SEE WHY IT'S HOT

Watch the original review

Go back to the video that brought you here.

Watch on TikTok ↗

Don't miss the next repo

Every new repo gets a video review, with its install guide right here.

Follow on TikTok ↗

HISTORY

Seen on GitHub Trending

  1. Sep 29#1061,030
  2. Sep 28#459,826
  3. Sep 27#658,728
  4. Sep 26#658,094
  5. Sep 24#156,173