Course Materials
Slides and video walkthroughs to accompany the book
This book is the written half of a course. The other half is a set of lecture slides and recorded walkthroughs, made for IESTI05 — Edge AI Engineering at UNIFEI and used in AIEng4D workshops. Everything is open and free to reuse under the same licence as the book.
- Slide decks (PDF): github.com/Mjrovai/UNIFEI-IESTI05-EDGE_AI
- Video playlist: Generative AI with Arduino UNO Q
- Weekly labs: one lab per class, with objectives, instructions, and a deliverable
What Maps to What
The lecture numbering is its own sequence and does not line up with the book’s chapter order, so use this table rather than the numbers. Lecture 1 is the UNO Q setup, which is the second chapter of Part 1 in the book.
Three chapters have no lecture yet. SLMs at the Edge runs entirely on a student’s own computer and works well as pre-class reading; the two Part 2 chapters cover the App Lab and Edge Impulse workflow, which is easier to demonstrate live than to record.
Total running time of the six recorded lectures is about 1 hour 38 minutes.
How the Course Runs
In IESTI05 this material occupies thirteen one-hour classes, numbered 16 to 28 in the course. An hour is enough to introduce a topic and start the lab, not to finish it, so students work through the corresponding chapter themselves between sessions, and the longer chapters are split across more than one class.
| Class | Topic | Chapter |
|---|---|---|
| 16 | How generative AI works — SLMs | How Generative AI Works |
| 17 | SLMs at the edge, on the student’s own machine | SLMs at the Edge |
| 18 | UNO Q setup walkthrough | Arduino UNO Q Setup |
| 19 | Running SLMs on the UNO Q | Generative AI at the Edge |
| 20 | Benchmarking and model selection | Generative AI at the Edge, §7 |
| 21 | Talking to the model from Python | Generative AI at the Edge, §9 |
| 22 | Multimodal AI at the edge | Multimodal AI at the Edge |
| 23 | SLM optimization — agentic AI | Agentic AI at the Edge |
| 24 | Building an agent on the UNO Q | Agentic AI at the Edge, §5 |
| 25 | Physical AI and Bridge RPC (MPU to MCU) | GenAI Meets the Real World |
| 26 | Physical AI project | GenAI Meets the Real World, §6 |
| 27–28 | RAG at the edge | in development |
The live plan, which is more current than this table, is in the course repository: IESTI05 — Arduino UNO Q. It also carries a parallel Raspberry Pi track drawn from the companion volume.
Note that the course takes the agentic chapter before the physical-AI one, reversing the book’s order. That works because Agentic AI at the Edge depends only on the setup and llama.cpp chapters — it says so explicitly in its own prerequisites, and it deliberately avoids sensors and vision so it can be taught early.
Two things worth copying if you run this yourself. Get the 4 GB UNO Q; the 2 GB board will not hold the models used from Class 19 onward. And put SLMs at the Edge before the boards are handed out, as Class 17 does, so nobody is idle waiting on hardware.
If you teach with this material, I would like to hear about it — corrections and contributions are welcome in either repository.