Edge Artificial Intelligence has moved, within the span of a single decade, from a specialized research curiosity to one of the defining engineering disciplines of contemporary computing. Billions of devices - industrial sensors, wearable health monitors, smart cameras, autonomous vehicles, agricultural equipment, consumer electronics - now carry some measure of on-device intelligence, and the pace of this deployment shows no sign of slowing. Yet despite this rapid, widespread adoption, the field has, until recently, lacked a body of instructional literature that treats it as the genuinely interdisciplinary subject it is: neither a narrow subfield of machine learning, nor purely an embedded-systems specialization, but a discipline that demands fluency across both, plus real-time systems theory and distributed systems engineering besides.
This book was written to fill that gap. It is intended for advanced undergraduate students, graduate students, and practicing engineers who already possess some grounding in either machine learning or embedded systems - but not necessarily both - and who need a single, coherent, technically rigorous treatment that connects the two. Throughout, the book deliberately privileges engineering trade-offs and system-level reasoning over any single technology's implementation details, on the conviction that specific hardware platforms and software frameworks will continue to evolve rapidly, while the underlying analytical framework for reasoning about latency, bandwidth, energy, memory, and real-time determinism will remain durably useful regardless of which specific technologies a reader ultimately works with.
The book's structure reflects this philosophy. Part I establishes shared foundations across machine learning and embedded systems, deliberately written to be accessible to readers strong in one domain but less familiar with the other. Part II develops the core technology stack - hardware accelerators, model compression, software toolchains, and real-time operating systems - that connects a trained model to a deployed, resource-constrained device. Part III surveys applications in substantial technical depth, from computer vision through industrial and smart-city deployments, always grounding the discussion in the foundational and technological material developed earlier. Part IV addresses the advanced and forward-looking concerns - security, federated learning, energy management, and emerging trends - that increasingly separate mature, production-grade edge AI engineering from early-stage prototyping.
Every chapter includes original diagrams designed specifically for this book to illustrate its core architectural and conceptual content, tables summarizing comparative technical trade-offs, worked case studies drawn from realistic deployment scenarios, chapter-ending review questions suitable for coursework or self-assessment, and a curated reference list pointing toward the foundational and current literature underlying each chapter's content. It is my hope that readers - whether working through this book cover to cover as a course text, or consulting individual chapters as a technical reference during a specific project - come away with both the conceptual vocabulary and the practical engineering judgment needed to design edge AI systems that are not merely accurate in isolation, but genuinely deployable: correctly sized to their hardware, defensible under real-time constraints, secure against realistic threats, and honest about their energy budget.
