TinyML
Historical materials · Volume I
Overview
Students train models and deploy them on microcontrollers, where limits on memory, energy, and latency shape every design choice. This page collects the materials from the Harvard CS249r offerings and the HarvardX TinyML program. The course assumes Python and basic linear algebra.
Syllabus and schedule
The TinyML syllabus pairs each week of the courseware with a Volume I chapter. The syllabi of past Harvard offerings appear under Offerings.
Materials
| Resource | Use |
|---|---|
| Lecture slides | Decks from the HarvardX TinyML courseware. |
| Hardware kits | Microcontroller boards and deployment labs. |
| Fall 2020 assignments | Starters for hello world, gesture recognition, person detection, wake words, and responsible AI. |
| Fall 2022 Assignment 2 | Visual wake words: training, quantization, and deployment (Colab). |
| TinyMLx courseware and Colabs | Structure, readings, and notebooks from the HarvardX specialization. |
| TinyMLx Arduino library | Source and examples for the TinyML kit and shield. |
For instructors
The historical code and hardware instructions have not been rerun on current toolchains, so each assignment needs a test run before it is assigned.
Offerings
| Offering | Course overview | Syllabus and schedule |
|---|---|---|
| Fall 2023 | Course website | Syllabus |
| Fall 2022 | Course website | Syllabus |
| Fall 2020 | Course website | Syllabus and schedule |
Project showcases from 2023 and 2022 remain on the original course site. Selected projects with public code:
- Cache 4 Trash: waste sorting with a Nicla Vision classifier and a motorized sorter.
- Activity recognition and ultrasound power control: activity recognition on a Nicla Vision that controls ultrasound power.
- Air Guitar: gesture recognition from motion and finger-flex sensors on a Teensy board.
The course directory lists TinyML courses at other institutions.
Links reviewed October 21, 2026.