Xiaofan Yu received the B.S. degree from Peking University, China in 2018 and the M.S. degree from University of California at San Diego in 2020. She is currently pursuing the Ph.D. degree with the Department of CSE, University of California at San Diego. She is working under the supervision of Prof. Tajana Rosing, while she has a history of successful collaborations with Dr. Ludmila Cherkasove (Arm Research), Prof. Yunhui Guo (UTDallas), Prof. Arya Mazumdar (UCSD) and Prof. Sicun Gao (UCSD). Her work has resulted in publications in top venus in the CPS area, e.g., ESWEEK, CPS-IoT week and MobiCom. She is expected to graduate in Spring 2024.
Along with the recent advancements of lightweight machine learning and powerful platforms (e.g., Raspberry Pi), learning at the edge has become the next tide of IoT. On-device training allows devices to learn directly from the data they gather, which dramatically reduces the total amount of data that has to be sent via network, thus reducing the overall energy costs. It also provides much faster decision making at the edge, without the need for constant connectivity to the cloud. Though promising, multiple challenges from data, algorithm, network and hardware aspects lie in the way of deploying intelligence at the edge. My research targets at addressing these barriers and closing the gap towards enabling intelligent and large-scale IoT deployments in the real world, which has three main thrusts: (1) adaptive and robust AI on a single embedded device, (2) efficient distributed learning in IoT networks and (3) brain-inspired lightweight AI with custom hardware. I am hoping to contribute a full-stack of technologies and demonstrate in real-world ecological applications, e.g., the High Performance Wireless Research and Education Network (HPWREN) in Southern California.