A master’s candidate Jaegeun Bae from Professor Jeongho Kim’s research group (KAIST TERA Lab) has won the Best Student Award at EDAPS (Electrical Design of Advanced Packaging & Systems) 2025, the most prestigious international conference on semiconductor packaging technologies in the Asia–Pacific region.
With this achievement, TERA Lab has produced award winners for two consecutive years, following last year’s Best Paper Award received by master’s student Tae-soo Kim, further solidifying its global research excellence.
At EDAPS 2025, held from December 15 to 17 in Sapporo, Japan, Bae presented his paper entitled “Switch Transformer-based HBM Design Agent.” Among more than 30 papers published during the year, his work was recognized for its significant contribution to technological innovation in the field and was selected as the Overall Best Student Paper Award (Best Student Award) at EDAPS 2025.
EDAPS is the largest and most influential international conference on semiconductor packaging technologies in the Asia–Pacific region. Since 2002, it has been annually organized and hosted by the IEEE Electronic Packaging Society. The conference brings together academic researchers and industry engineers primarily from the electrical engineering community and is widely known for providing a platform to share research outcomes and conduct industry-oriented studies across a broad range of topics, including chip design, System-in-Package (SiP) and System-on-Package (SoP), electromagnetic interference and compatibility (EMI/EMC), electronic design automation (EDA) tools, 3D-IC, and through-silicon via (TSV) design.
Each year, on the final day of the conference, EDAPS announces award-winning papers in three categories: Best Paper Award, Best Student Award, and Best Poster Award, selected from papers submitted that year.
Bae’s paper, “Switch Transformer-based HBM Design Agent,” applies a switch transformer-based reinforcement learning algorithm to suppress Power Supply Induced Jitter (PSIJ)—a major cause of signal integrity degradation—below a target threshold while minimizing the number of decoupling capacitors. The proposed approach demonstrated approximately 15% improvement in inference speed compared to conventional optimization algorithms, drawing significant attention from the community.
In particular, the paper was highly praised by the review committee for presenting a novel methodology to address the increasingly shrinking PSIJ margin in High Bandwidth Memory (HBM) caused by rising data rates. Moreover, Bae proposed an original system with high reusability, applicable not only to current HBM designs but also to future-generation and next-generation HBM architectures, which contributed to the paper’s strong evaluation.
Bae commented, “Professor Jeongho Kim—often referred to as the ‘father of HBM’—provided invaluable guidance throughout the process of systematically organizing the theme and content of this research,” adding, “I hope this work will serve as a small but meaningful first step toward establishing agentic AI that integrates both hardware and software design for HBM, which is the direction currently pursued by TERA Lab.”
He further stated, “Beyond PSIJ optimization, I aim to expand this research into a full-lifecycle HBM design agentic AI that comprehensively considers power and signal integrity as well as thermal characteristics,” and expressed his aspiration “to build a practical AI-based design framework applicable to next-generation HBM and chiplet-based architectures, ultimately contributing to real-world industrial applications.”
As of December this year, TERA Lab consists of 27 students, including 18 master’s and 9 doctoral candidates, who are actively conducting research on optimizing various front-end and back-end semiconductor packaging and interconnection designs using artificial intelligence and machine learning techniques, such as reinforcement learning and imitation learning.
In addition to Bae’s recent award, TERA Lab continues to receive global recognition in semiconductor design research. Earlier this year, Tae-soo Kim, a master’s graduate who is now pursuing a Ph.D. at Georgia Institute of Technology, won the Overall Best Paper Award at EDAPS 2024. Furthermore, doctoral student Tae-in Shin received the Best Paper Award at DesignCon, another internationally renowned conference, underscoring TERA Lab’s world-class research capabilities in the field of semiconductor design.
