Enabling Resource-Efficient AI via SW-HW Co-design

The NSF AI Institute for Edge Computing (Athena) is pleased to present the next in the Seminar Series by Jong Hwan Ko, Kang Eun Jeon, Johnny Rhe, and Sangheum Yeon, titled "Enabling Resource-Efficient AI via SW-HW Co-design" on Wednesday, August 13, 2025, from 3-5pm EST in-person at Duke University Wilkinson Building Room 132, and via Zoom. Refreshments will be served. Abstract: The rapid growth of AI model size and complexity demands efficient deployment on resource-constrained systems. This challenge calls for a principled co-design of SW and HW, as conventional approaches often fail to exploit the joint optimization opportunities. In this seminar, we explore a range of SW-HW co-design techniques geared towards resource-efficient AI across a wide range of real-world applications. First, we present a series of model compression techniques, with a focus on mixed-precision and multi-precision quantization to accelerate both inference and training. Next, we explore how compression techniques can be co-optimized with next-generation computing paradigms (e.g., in-memory computing) to enhance computational efficiency. Through this seminar, we aim to offer practical pathways for efficient AI deployment under resource constraints.