Course overview
This course is a practical guide to the Edge AI workflow, moving from hardware selection to fully deployed machine learning models on microcontrollers. It begins with a detailed tutorial on an x86 development board before diving into the core of on-device inference with TensorFlow Lite model conversion and quantization. You will then explore the TinyML pipeline using Arduino, ESP32, and STM32 platforms, covering sensor data collection, neural network training, and C++ firmware integration. The curriculum also includes specialized sessions on deploying computer vision with CNNs and leveraging STM32’s X-CUBE-AI expansion pack for optimized performance.
Who this course is for
This course is for embedded systems engineers, IoT developers, and software programmers with basic Python or C++ knowledge who want to implement AI locally on hardware. It suits builders working with Arduino or ESP32 who need to run models without cloud connectivity. If you face challenges optimizing models under tight memory and processing constraints or want hands-on experience deploying real-time classification and object detection on microcontrollers, this sequence provides a direct, code-focused path.
How to study this sequence
Follow the lessons sequentially to build a complete understanding of the Edge AI deployment pipeline. Start with the x86 hardware tutorial to grasp embedded constraints, then move into TensorFlow Lite optimization and quantization. As you progress into TinyML, run the provided code examples on your own Arduino, ESP32, or STM32 boards. Pause to replicate the training workflows in Google Colab or Edge Impulse before attempting the firmware deployment steps. Ensure each model runs locally on your physical hardware to validate your understanding of inference latency and memory usage.
What you should be able to do
By completing this course, you will be able to collect sensor data, train a neural network, and deploy it as an optimized model directly onto a microcontroller. You will gain practical skill in using TensorFlow Lite, Edge Impulse, and STM32Cube.AI to implement real-time inference for applications like gesture recognition and object detection. The final result is a working embedded system performing AI tasks on-device, ready for you to adapt to custom hardware projects.
