Afghanistan Artificial Intelligence Institute
Build intelligence.
Transform Afghanistan.
A free AI school and technology community. Different Artificial Intelligence subjects, in order — Python and mathematics through to generative AI, agentic systems and MLOps — taught properly, so that Afghan engineers build these systems instead of waiting for them.
The pattern behind this page is a gul — the octagonal medallion of Turkmen and Afghan carpets — woven knot by knot, the way a weaver would set it. Every carpet is a grid; every grid is a matrix. Afghanistan has been computing patterns for six thousand years.
Our mission
Talent is everywhere. Teaching is not.
We are building a complete AI education and putting it within reach of people who have been left out of one — free, open, and designed to work on a slow connection and a modest laptop.
Free, and free means free
No fees, no locked modules, no premium tier, no advertising. Everything published under an open licence you can teach from.
On real data
Exercises built on real data — instead of the same three textbook datasets everyone else uses.
From zero to shipping
One continuous path from your first line of Python to a monitored system in production. Not a pile of unconnected videos.
A community, not a playlist
Study groups, questions answered, mentors and group projects. Learning alone is hard, and most people stop.
Curriculum
Ten subjects, in order
This is a sequence, not a menu — each subject depends on the one before it. Everything is in production now, and we publish a subject only when it is complete rather than releasing half of it.
Python
The language the rest is written in. Install it, break it, fix it, then read and write real programs.
Mathematics for AI
Linear algebra, calculus and probability — the parts that actually appear inside models, and an explicit list of what you can skip.
Machine Learning
Regression, classification, trees and boosting. Training and validation, overfitting, and choosing a metric that is not misleading.
Deep Learning
Backpropagation by hand, then neural networks, then attention and the transformer architecture behind every model you have heard of.
Computer Vision
Images as numbers, convolution, and classifiers that survive photographs taken by a real phone in bad light.
Natural Language Processing
Tokenisation, embeddings and language models — including what breaks on low-resource languages, which is most of them.
Generative AI
Prompting, fine-tuning, retrieval-augmented generation and honest evaluation. Making a model answer from your documents, with citations.
Agentic AI
Tool use, planning loops, memory and multi-agent structure. Build an agent that does a real job, then prove that it works.
MLOps
The part most courses skip: serving behind an API, containers, monitoring, drift, cost, and rollback at three in the morning.
End-to-End Projects
One complete system, start to finish, on a problem that matters. This is the portfolio piece an employer actually reads.
Video lessons
Video lessons are being recorded
Short and practical, free on YouTube, downloadable so you can watch offline, and available at 360p for slow connections. The first set covers Python and mathematics.
Python, from installation to your first program
Subject 01 · in production
The linear algebra you actually need
Subject 02 · in production
What a neural network is, explained with a carpet
Subject 04 · in production
Retrieval-augmented generation, end to end
Subject 07 · in production
Who this is for
Four kinds of people, one curriculum
We have no graduates yet, so there is nothing here we can honestly call a success story. What we can tell you is who we built this for.
Students
School and university students who want a real technical skill rather than a certificate. Start at subject 01 with no prior knowledge and no laptop required for the first weeks.
Teachers
Teach any of this in your own classroom. Slides, lesson plans and marking guides are free and openly licensed, and we will train your teachers over a weekend.
Professionals
Developers and analysts moving into machine learning. Skip to subject 03, and use MLOps and the end-to-end project to build something you can show an employer.
Organisations
NGOs, clinics, companies and agencies with a real data problem. Bring it to us as a supervised student project and keep the result.