Let’s Talk Technical: Edge AI Part 2, From Smart Devices to Physical AI | DigiKey

Edge AI is moving beyond connected devices and intelligent sensing, but what is driving adoption, and what does it take to build AI systems that can actually act on the physical world? Join DigiKey moderator Shawn Luke alongside technical experts Shawn Hymel, Aashish Chaddha from Silicon Labs, Marcus Mayr from STMicroelectronics, Davis Sawyer from NXP, and Nilam Ruparelia from Microchip Technology as they examine the evolution of Edge AI, from smart consumer devices and embedded intelligence to physical and embodied AI. The panel explores why simply connecting devices is no longer enough. As AI-capable hardware, accelerators, models, and development tools continue to mature engineers can increasingly process data where it is generated, reducing latency, bandwidth requirements, cloud costs, and privacy concerns while enabling new capabilities in embedded systems. The discussion covers real-world examples including smart doorbells, thermostats, industrial safety systems, sensor monitoring, computer vision, robotics, autonomous mobile robots, and other intelligent edge applications. The panel also examines how AI can create new business opportunities by turning device intelligence into practical services and capabilities. Topics include: • Why Edge AI adoption has accelerated in recent years • Connected devices vs. truly intelligent devices • Real-world Edge AI applications and use cases • On-device AI, latency, bandwidth, privacy, and cloud costs • Choosing Edge AI hardware beyond TOPS performance • CPU, memory, peripherals, power, and thermal constraints • Reliability and long-term product availability • SDKs, documentation, developer experience, and technical support • Time-to-prototype and time-to-market considerations • Software portability and the role of Zephyr RTOS • Avoiding vendor lock-in during AI development • AI hardware designed for consumer vs. industrial applications • Generative AI and multimodal systems at the edge • Physical AI and embodied AI The conversation also examines an important evolution in Edge AI. Earlier generations of embedded AI primarily focused on perception, detecting an object, recognizing a person, identifying an anomaly, or interpreting sensor data. Physical AI takes that concept further by combining perception with decision-making and physical action, enabling systems such as robotic arms, autonomous mobile robots, and other machines to respond directly to their environments. For developers, this shift creates new opportunities as well as new challenges. Systems that interact with the physical world require greater compute capacity, stronger safety considerations, reliable hardware, and software architectures capable of handling increasingly complex AI models. Whether you're developing industrial IoT, smart sensors, robotics or next-generation AI products, this discussion provides practical insight into selecting the right hardware, software, and development approach for deploying intelligence at the edge.