Accelerators for Edge AI: From Architecture to Applications

The growing demand for intelligent edge devices is driving the development of specialized AI accelerators capable of delivering high performance under strict power and resource constraints. Key design challenges and trade-offs in this domain relate to energy efficiency, performance, memory bandwidth, and hardware-software co-design.

Drawing on recent research activities at SDU Microelectronics, domain-specific hardware can enable real-time machine learning directly on edge devices without relying on cloud connectivity. Practical application areas include intelligent sensing, healthcare monitoring, industrial automation, and Internet-of-Things (IoT) systems.

In this talk, Farshad Moradi will provide an overview of AI accelerator architectures for edge computing and discuss how SDU’s research addresses key hardware trade-offs. The presentation will cover application examples across healthcare and industrial automation, concluding with perspectives on future trends and opportunities for innovation in edge AI hardware.