AI Magic Wand with Arduino Part 2: Training our Wand to Recognize Spells
Train, test and deploy your TinyML model from Edge Impulse to Arduino.
In Part 1, we taught our wand what each spell looked like by collecting over 200 gesture recordings. Now it’s time to transform all of that data into a real machine learning model.
In this session, we’ll walk through the entire TinyML workflow using Edge Impulse. You’ll learn how to extract features from your recordings, train a neural network, evaluate its accuracy, and deploy the finished model back onto your Arduino Nano 33 BLE Sense.
Then comes the exciting part…
We’ll put our AI to the test by casting real spells! Using the Serial Monitor, you’ll watch your wand recognize gestures in real time with an example sketch conveniently bundled into an Arduino library by Edge Impulse.
Along the way, you’ll learn:
- How Edge Impulse transforms raw sensor data into machine learning features
- Training and evaluating a TinyML neural network
- Understanding accuracy scores, confusion matrices, and model performance
- Deploying your trained model back to the Arduino Nano 33 BLE Sense
- Running real-time inference directly on your microcontroller
- Responding to recognized gestures with Arduino code
By the end of the session, your wand won’t just collect motion data—it will recognize your gestures completely offline and respond instantly without needing a computer or internet connection.
Whether you’re building props for Halloween, cosplay, escape rooms, theater, or simply curious about TinyML, this workshop will give you the skills to build intelligent projects that react to the world around them.
Because once your Arduino starts recognizing your movements, you’re no longer programming every action…
You’re teaching it to recognize them.
Code & Build Along With Me
Affiliate Disclosure: Some links here might earn me enough nuts and bolts (aka commission) to keep the LEDs blinking and servos turning. No extra cost to you, just a lil’ fuel for my caffeine-powered nights of DIY-ing. I only present stuff that’s survived my “Did-it-Explode?” test. Build on, and thanks for the recharge!
Important: This guide assumes you’re using the Arduino Nano 33 BLE Sense Rev2. The Rev2 includes an updated IMU sensor, and the software, libraries, and firmware demonstrated throughout this workshop are based on this newer hardware revision.

You’ve Reached the Workshop Door..
The video walkthrough got you this far.
Inside Burnt Breadboard Society, you’ll unlock the complete Project Guide for this build, including everything you need to recreate it with confidence—without constantly pausing, rewinding, or guessing where that one mysterious wire goes.
Inside you’ll unlock:
- Step-by-step Project Guide
- Detailed wiring diagrams
- Copy-and-paste Arduino & ESP32 code
- Complete parts list with links
- STL files (when applicable)
- PCB files (when applicable)
- Full workshop replay
- Troubleshooting & common fixes
- Access to our private community and live Open Lab sessions for personalized help
Whether you’re building spooky props, robots, cosplay electronics, escape room puzzles, or something completely original, Burnt Breadboard Society gives you the tools, guidance, and support to build projects you’ll actually be proud to show off.
Already a member?
LOG IN here to access this Project Guide instantly.
