Physical AI at the Edge: Autonomous Drones, Embedded Vision, and Hardware-Optimized Learning Edge AI is redefining how intelligent systems operate in real-world, bandwidth-limited, and safety-critical environments. Rather than relying on centralized cloud infrastructure, next-generation autonomous platforms demand low-latency, energy-efficient, and hardware-aware intelligence deployed directly at the edge. This guest lecture presents a systems-level perspective on designing edge-native AI architectures spanning three converging domains: autonomous drone platforms, embedded computer vision, and lightweight large language models (LLMs). We will examine how drone autonomy benefits from onboard perception-action loops, real-time multi-modal sensing, and closed-loop control frameworks that operate without persistent cloud connectivity. The talk further explores edge-optimized vision pipelines for agricultural and environmental monitoring, including model compression, quantization, spectral-RGB fusion, and real-time deployment on resource-constrained hardware such as Jetson, FPGA, and heterogeneous SoCs. Finally, we discuss the emerging role of LLMs in physical AI systems, not as chat interfaces, but as structured reasoning engines integrated with robotic sensing and decision pipelines. Hardware-aware optimization strategies, including pruning, mixed-precision inference, memory- latency tradeoffs, and accelerator-centric design, will be highlighted as key enablers of scalable deployment. This invited talk concludes by outlining a unifying design framework for building autonomous, interpretable, and deployable edge intelligence systems across agriculture, environmental monitoring, and cyber-physical domains. All are welcome! You do not need to be a member to attend. If you are interested and unable to attend, please register and a recording will be sent out after the event. Speaker(s): Prabha , Virtual: https://events.vtools.ieee.org/m/531954
