The AI Engineering Blueprint

A course in a box for teaching AI systems. Two semesters of material, open source and ready to adopt.

Students need Python proficiency, matrix multiplication and transposes, and introductory probability. No prior ML coursework. More →
You need No GPUs and no cluster. Labs run in the browser; hardware kits are optional and the course runs fully without them. More →

Syllabi

Two-semester course timeline showing four parts per semester with key milestones.

Course Materials

Textbook Two volumes: Foundations (1–8 GPUs) and At Scale (distributed fleets). HTML, PDF, EPUB. Vol I · Vol II
TinyTorch Students build a framework from scratch, tensors through transformers, across 20 modules. The tests grade themselves. Browse →
Interactive Labs Run in the browser on mlsysim with no GPU and nothing to install, so lab access costs nothing to provision. Browse →
Hardware Kits Arduino Nano 33 BLE, Raspberry Pi with Coral, Seeed XIAO ESP32S3. Optional; the course runs fully without hardware. Browse →
Lecture Slides One deck per lecture, each with speaker notes and in-class exercises. PDF and PowerPoint, or edit the LaTeX source. Browse →

Teaching Resources

Syllabi Week-by-week schedules with linked readings, labs, and assignments for both semesters. Sem 1 · Sem 2
Assessment Three-tier rubrics, sample student work, AI Olympics capstone spec, grading load estimates. View →
Pedagogy Prediction Locks, Decision Logs, the A→B→C lab structure, and the Iron Law audit framework. View →
TA Guide Grading workflows, common student struggles by week, lab facilitation, office hours protocol. View →
Customization 10-week quarter, 3-day workshop, graduate seminar, embedded/cloud emphases. View →
FAQ Prerequisites, setup, AI tools policy, hardware budgets, and adoption questions. View →

The Books

The two volumes students read. Free online in full, with hardcover editions published by MIT Press.

Machine Learning Systems, Volume I: Foundations hardcover edition
Semester 1 · Volume I

AI Systems Foundations

One machine to eight accelerators. The Iron Law of hardware-software co-design and the full stack, built from tensors up.

Machine Learning Systems, Volume II: At Scale hardcover edition
Semester 2 · Volume II

AI Engineering at Scale

One node to ten thousand. Distributed training, collective communication, and fleet infrastructure for frontier models.

Companion Books

The MLSysBook curriculum is extended by open companion books authored by members of the TinyML4D Academic Network. These pair directly with the hardware kits used in Semester 1's labs, and are maintained by the original authors.

Ready to Adopt?

Part of the MLSysBook Ecosystem

Back to top