Four KAIST PhD Candidates Simultaneously Win Best Paper Award at 'DesignCon'

[Asia Economy (Daejeon) Reporter Jeong Il-woong] Four KAIST doctoral students simultaneously won the Best Paper Award at the international semiconductor design conference, DesignCon.

KAIST announced on the 16th that doctoral candidates Kim Sung-guk (31), Choi Sung-wook (27), Shin Tae-in (26), and Kim Hye-yeon (female, 26), who are studying at the Terra Lab in the Department of Electrical Engineering, were selected as recipients of the '2022 Best Paper Award' at DesignCon.

(From left) KAIST Department of Electrical Engineering Kim Seong-guk, Choi Seong-wook, Shin Tae-in, Kim Hye-yeon PhD candidates. Provided by KAIST

DesignCon is an internationally recognized conference with global authority in the semiconductor design field, and KAIST demonstrated its prowess by producing four of the eight total awardees at this conference.

KAIST explained that their Best Paper Award wins are especially meaningful as they were achieved in competition with engineers and researchers from global big tech companies in semiconductor powerhouses such as the United States, China, and Japan (including Intel, Micron, AMD, Huawei, etc.).

The awarded papers concern semiconductor design using artificial intelligence (two papers) and semiconductor architecture design for AI computing (two papers).

Kim Sung-guk designed a high-bandwidth memory-based processing-in-memory (PIM) architecture for high-performance AI accelerators, and Choi Sung-wook attracted attention by designing a hybrid equalizer for high-bandwidth memory (HBM) using reinforcement learning methodologies.

Additionally, Shin Tae-in proposed signal integrity modeling, design, and analysis methodologies for next-generation neuromorphic computing systems, and Kim Hye-yeon defined the decoupling capacitor placement problem in semiconductor design as a combinatorial optimization problem and automatically optimized it through imitation learning, an offline learning method.

In particular, Kim Hye-yeon's research was recognized for designing learning methods and neural architectures tailored to the characteristics of semiconductor design problems and was also presented at the NeurIPS workshop, the largest AI conference held in early 2022.

The Terra Lab, to which these awardees belong, researches next-generation AI semiconductors such as HBM (high-bandwidth memory) used in high-performance servers for supercomputers and large-scale data centers by combining reinforcement learning (RL) and machine learning (ML) technologies that enable AI to autonomously implement optimal designs, along with 3D heterogeneous semiconductor packaging technology.

Their research is gaining attention for leading the development of AI semiconductors and computers that enable digital transformation centered on artificial intelligence, presenting a future direction that automates the entire design process through AI.

This year’s award ceremony will be held at the 'DesignCon 2023 International Conference' at the Santa Clara Convention Center in San Jose, Silicon Valley, USA, on the 31st of this month.

Professor Kim Jeong-ho, advisor of Terra Lab, said, "In the era of digital transformation (DX), the role of semiconductors is becoming increasingly important," adding, "We will strive to nurture many customized talents necessary for next-generation semiconductor development at Terra Lab."

Meanwhile, DesignCon is held annually targeting researchers and engineers from global big tech companies such as Intel, Micron, Rambus, Texas Instruments (TI), AMD, Huawei, IBM, ANSYS, as well as university and graduate students worldwide.

This content was produced with the assistance of AI translation services.