Level 1 · Entry Foundations¶
Goal: understand what edge AI is and master the complete TinyML workflow — train a small model with Keras, convert it to TensorFlow Lite (LiteRT), quantize it to int8, turn it into a C array for TFLite-Micro, and understand exactly what changes when it runs on a microcontroller — finishing with a complete gesture-recognition pipeline.
You do not need to own any hardware for this level. Every module runs on a normal laptop with Python and TensorFlow: training, conversion, quantization, and even the C++ inference code (which compiles as a desktop program before it ever touches a board). Microcontroller deployment is covered as the follow-on step in each module, with Wokwi simulator notes where the ESP32 side can be tried in the browser.
Modules¶
- What Is Edge AI
- The Edge AI Workflow
- Training a Tiny Model
- Converting to TFLite/LiteRT
- Quantization
- TFLite-Micro & C Arrays
- Deploying to a Microcontroller
- Sensors & Feature Extraction
- Evaluating & Debugging Edge Models
- Capstone — Gesture Recognition
By the end of this level you'll be able to take a machine-learning idea from
a Keras training script to a quantized .tflite flatbuffer to a C array
compiled into microcontroller-style firmware — verifying at every step that
the model still produces the right answers — and you'll know precisely what
RAM, flash, latency, and accuracy each step costs.