ABOUT
From a classroom
to six continents.
The story of an open-source curriculum that grew into a global movement for AI engineering education.
Our Story
In 2020, a new graduate course launched at Harvard: CS 249r, focused on the emerging field of TinyML — machine learning on the smallest, most constrained devices. There was no textbook for it. So we started writing one.
The first version was rough — course notes stitched into chapters, shared with students as a PDF. But when we open-sourced it on GitHub, something unexpected happened. Students and educators around the world started reading it, contributing to it, and teaching from it.
In 2021, the International Centre for Theoretical Physics (ICTP) in Trieste partnered with us to form the TinyML4D Academic Network — 40+ universities across six continents, each receiving curriculum materials and hardware kits. What started as a Harvard course was now taught in Bogotá, Nairobi, Kobe, and Johannesburg.
By 2024, the material had outgrown TinyML. Students were asking about training at scale, distributed systems, fleet orchestration — the full systems stack. The textbook expanded into two volumes: Introduction to Machine Learning Systems and Machine Learning Systems at Scale. The website became mlsysbook.ai. MIT Press signed on for a hardcover edition.
Today the project is more than a textbook. It is a full curriculum — with interactive labs, a build-from-scratch ML framework (TinyTorch), lecture slides, hardware kits, instructor resources, and an interview prep guide. All of it open. All of it free.
Mission
The world is rushing to build AI systems. It is not engineering them.
Machine learning has transformed every industry, but the education pipeline has not kept pace. Students learn algorithms in isolation, divorced from the silicon that executes them, the memory hierarchies that constrain them, and the networks that distribute them. The result: a generation of practitioners who can call model.fit() but cannot reason about why their model is slow, where the bottleneck lives, or how to fix it without guessing.
Machine Learning Systems exists to close that gap. We teach AI as infrastructure — governed by physical laws, shaped by engineering constraints, and measurable through quantitative reasoning. Our thesis is simple: constraints drive architecture. You do not choose a Transformer because it is trendy; you choose it because of how it parallelizes on real silicon.
⚙️
Principles Over APIs
We teach the physics beneath the framework. Memory hierarchies, not torch.cuda(). Arithmetic intensity, not library calls. Knowledge that endures beyond any single tool.
📐
Quantitative Reasoning
Every claim is backed by a number. Students learn to estimate, measure, and verify — the same skills that separate a systems engineer from a script runner.
🌍
Open and Global
Free HTML, PDF, and EPUB. Used at universities across six continents. Licensed under CC-BY-NC-SA 4.0 so anyone can teach from it, adapt it, and build on it.
🔬
Full-Stack Curriculum
Not just a textbook. TinyTorch, hardware kits, interactive labs, lecture slides, and instructor resources — everything needed to teach a two-semester course.
22,800+
GitHub Stars
2
Volumes
32
Chapters
95+
Contributors
6
Continents
Milestones
2020
CS 249r launches at Harvard
The first offering of the graduate course on TinyML systems. Course notes become the seed of the textbook. The edX TinyML Professional Certificate launches with Google and the tinyML Foundation.
2021
The TinyML4D network forms
Partnership with ICTP in Trieste creates the TinyML4D Academic Network. First cohort of 20 universities joins from Latin America, Africa, and Asia. SciTinyML workshop series begins.
2022
Open-source textbook goes live on GitHub
The textbook is published as an open repository. Community contributions begin. Second cohort of 20 universities joins TinyML4D. Stars grow as the repo gains visibility.
2024
Beyond TinyML — Machine Learning Systems
The textbook expands from TinyML into the full ML systems stack: training, serving, distributed systems, fleet orchestration. Two volumes take shape. mlsysbook.ai launches. IEEE formally cites the work.
2025
The full curriculum takes shape
TinyTorch (build-from-scratch ML framework), interactive Marimo labs, hardware kits with Arduino and Seeed, 35 Beamer lecture decks, and a complete instructor hub — the ecosystem goes far beyond the textbook.
2026
MIT Press publication
Two-volume hardcover edition forthcoming from MIT Press. 22,800+ GitHub stars. 95+ contributors. 40+ universities across six continents. The open-source edition remains free — always.
License
Machine Learning Systems is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC-BY-NC-SA 4.0).
This means you are free to share and adapt the material for non-commercial purposes, as long as you give appropriate credit and distribute your contributions under the same license.
CC-BY-NC-SA 4.0
Get involved
Whether you are a student, educator, researcher, or engineer — there is a place for you in this community.