Professor Seong-Lyun Kim - Professor Seong-Lyun Kim, School of EEE, Yonsei University, Seoul, Korea Communication systems
Date: - Location: Eurecom
Abstract: In this talk, I will introduce a GPU-accelerated vRAN testbed based on NVIDIA Aerial and OAI running on a DGX Spark (or H100 GPU) platform. The platform accelerates Layer 1 processing on the GPU and operates in the n79 band using South Korea’s 4.7 GHz private 5G spectrum. In a separate implementation, our team utilized the CUDA-accelerated OCUDU Layer 1 codebase as the gNB software baseline to establish end-to-end private 5G connectivity through a commercial indoor/outdoor O-RU. I will provide an overview of our benchmarking results across different combinations of open-source platforms—including OAI, OCUDU, and Aerial—and discuss changes in CPU utilization enabled by GPU acceleration while quantifying the associated processing overhead. Furthermore, I will present our latest results on hybrid inference combining SLMs (on terminals) and LLMs (on GPU vRAN) to investigate AI-RAN use cases, where the GPU simultaneously executes real-time RAN processing and AI inference workloads. Finally, I will provide an overview of the ongoing research and development initiatives for AI-Native Networks in South Korea. Bio: Seong-Lyun Kim is a Professor at the School of Electrical & Electronic Engineering at Yonsei University, Seoul, Korea, where he previously served as the Head of the School. He currently heads the Robotic & Mobile Networks Laboratory (RAMO) and the Center for vRAN Research (CVR). He leads the SNS JU EU-KR 6G ARROW project, co-directed the H2020 EU-KR PriMO-5G project, and currently oversees the campus-wide Private 5G AI-RAN deployment at Yonsei University. Previously, he was an Assistant/Associate Professor of Radio Communication Systems (Radiosystemteknik) at the Department of Signals, Sensors & Systems at the Royal Institute of Technology (KTH) in Stockholm, Sweden. He has published numerous papers in the area of wireless networks and co-authored the book, Radio Resource Management for Wireless Networks (with Prof. Jens Zander). He served as editorial member of IEEE, and Elsevier journals in the area of wireless networks and robotics, More recently, he was a co-recipient of the IEEE Heinrich Hertz Award for the research on federated learning in wireless networks.
