🤖 Embedded AI: Intelligence at the Deep Edge

Authored by David Such | Published by No Starch Press

Bringing artificial intelligence to the "deep edge" requires more than just a trained model. It demands a holistic engineering approach. This comprehensive guide bridges the gap between high-level machine learning and the gritty reality of hardware constraints.

---

📚 Book Specifications & Ordering

Note: Every print book purchased via nostarch.com includes a paid FREE ebook edition! Additionally, Early Access allows you to dive into full chapters months before the official release.

---

🛠️ The Engineering Challenge

Many developers assume that AI is simply about training a model. However, true embedded AI involves a complex pipeline:

Just train a model $\rightarrow$ The Real Process:

Drawing on over three decades of professional embedded engineering, David Such guides you through these decisions. Whether you are dealing with a flaky breadboard, noisy sensor inputs, or a tensor arena that is too small for your model, this book provides the how and the why for the fix.

---

🚀 What You Will Build

No prior machine learning experience is necessary. You will complete over 25 practical projects, supported by downloadable PCB designs, schematics, datasets, and source code.

• Voice-Activated Wake-Word Detector: A system that triggers on specific vocal commands.
• Real-Time Noise Suppressor: AI-driven audio cleaning.
• AI MIDI Synthesizer: A creative tool that composes original music.
• Self-Training Battery Monitor: A device that gathers its own training data.
• Neural Network Person Detector: A vision-based system running on a camera board.

---

📋 Prerequisites

To follow along with the projects, you will need the following hardware and software:

Hardware

• Primary Boards: Arduino UNO or Raspberry Pi Pico.
• Specialized Gear: A few projects require specific hardware boards (detailed in the text).

Software Stack

• Python (including the TensorFlow library).
• Arduino IDE.
• Raspberry Pi Pico SDK.

---

📖 Table of Contents

Part I: Foundations of Embedded AI

1. Path to Embedded AI
2. The Basics of Embedded Systems
3. Applied Machine Learning in Embedded Projects
4. Deep Learning $\rightarrow$ Understanding the math, such as the basic activation function: $f(x) = \frac{1}{1 + e^{-x}}$
5. Exploratory Data Analysis

Part II: Working with Sensors and Data

6. Smart Sensors
7. IMU Data Preprocessing
8. Sensor Fusion

Part III: Building Complete Systems

9. Sensor Machine Learning
10. Real-Time Audio Noise Suppression
11. Build an AI MIDI Synthesizer
12. Build a Hot Word Detection
13. Build an Arduino Battery Monitor and Logging Shield

Appendices: Epilogue, Project List, Resources, and Index.

---

👨‍💻 About the Author: David Such

David Such is a veteran embedded systems engineer and the founder of Reefwing Software, where he specializes in:

• IoT Device Development
• Robotics Platforms
• Drone Flight Control Systems

With a career spanning 30+ years, David has held senior positions at industry leaders including Honeywell, Tyco, and Serco Australia. He is a recognized voice in the "edge AI" community, providing technical insights for hardware engineers worldwide.

---

🖼️ Related Resources

Below are some other titles you might find useful for your technical library:

# Example: A conceptual snippet of what you might see in the book
def deploy_model(model, hardware_target):
    if hardware_target == "Pico":
        print("Optimizing tensor arena for Raspberry Pi Pico...")
        return model.quantize().convert_to_c_array()
    else:
        print("Deploying to standard Arduino...")
        return model.compress()