North Carolina State University, Bokyung Kim 교수님의 세미나가 7월 8일에 진행될 예정입니다.

관심 있는 학생 여러분들의 많은 참석 바랍니다.

■일시: 2026. 7. 8 (수요일) 14:30 ~ 16:00

■장소: 반도체관 400102호

■주제: “Beyond Intelligent Computing: Memory is the Center for Scalable, Efficient, and Trustworthy AI”

■연사: Bokyung Kim 교수님 (North Carolina State University)

Bokyung Kim is an Assistant Professor in the Department of Electrical and Computer Engineering (ECE) at North Carolina State University. She received her Ph.D. in Electrical and Computer Engineering from Duke University in 2024. Her research focuses on memory-centric design with full-stack co-optimization with broad experience in hardware design, spanning different hardware levels through device modeling, mixed-signal VLSI, architecture/system design, and real implementations in silicon. She received the Best Paper Award from the IEEE/ACM International Symposium on Low Power Electronics and Design in 2025 and was a selected fellow of the IEEE Laureate Forum in 2026, ACM Heidelberg Laureate Forum Young Researchers in 2024, and EECS Rising Stars in 2023. She serves as a member of Board of Governors in the IEEE Circuits and Systems Society, an Associate Editor TCAS-II, as a Secretary of the Machine-Learning CAS Technical Committee.

■ABSTRACT

Today’s AI technologies are increasingly constrained by compute-centric designs, leading to severe data-movement bottlenecks. These bottlenecks become especially acute as AI systems scale in model size and operate on sensitive data in privacy-critical domains. To enable sustained AI with scalability, efficiency, and trust, my research addresses this challenge by rethinking memory as an active computational fabric, rather than a passive storage element. In this talk, I will highlight the innovation and continued challenges of the memory-centric research in full-stack design approaches.

Specifically, the talk begins with newly developed 3D-memory device and circuit primitives for future scalability. Building on these 3D designs, I will introduce novel architectural paradigms that break traditional efficiency limitations, along with their silicon demonstrations for clinical epilepsy applications. Lastly, I will present our latest research on privacy-preserving system designs tailored for differentially private AI training. This talk will be concluded with future research directions toward trustworthy AI systems.

Through cross-layer co-design spanning circuits, architectures, and algorithms, memory-centric paradigms can move beyond intelligent computing toward sustainable and safe deployment in the real world.

■HOST: 고종환교수 (전자전기컴퓨터공학과)