TinyML
Overview
Deploying and optimizing ML models on microcontrollers with a focus on speed and efficiency.
A microcontroller focused ML project implementing quantized operators and lightweight inference loops for extremely resource constrained devices. Includes latency and energy profiling to evaluate performance trade-offs between accuracy, model size, and runtime efficiency. Features deployment ready templates for Arduino Nano 33 BLE Sense and similar boards.
Features
- Quantized kernels tailored for MCUs
- Model size reduction and calibration
- Latency and energy profiling tools
- Board specific deployment templates
AI on a Breadboard
This is my TinyML project in action. I successfully deployed both voice detection and image recognition models onto an Arduino Nano 33 BLE Sense. The setup looks basic, but the magic is in the chip. Each of those glowing LEDs corresponds to a different output from the AI confirming a voice command or identifying an object. It's awesome to see complex AI running on such low-power hardware.
