Gyeongchan Yun

Welcome to my page!

I am a research engineer at Kakao, working on optimization for training Large Language Model (LLM).

Previously, I earned my Ph.D. in computer science and engineering at UNIST, advised by Prof. Young-ri Choi, and received the B.S in computer science and engineering at UNIST. My research focuses on scalable distributed systems and GPU parallelism for large-scale AI model training.

rugyoon@gmail.com

github

linkedin

Affiliation

  • LLM research engineer at Kakao Corp, Oct. 2025 - Present

Education

  • Ph.D in Computer Science and Engineering at UNIST, Mar. 2019 - Aug. 2025 (Advisor: Prof. Young-ri Choi)
  • B.S in Computer Science and Engineering (Major) and Mathematical Science (Minor) at UNIST, Mar. 2014 - Feb. 2019

Publications

  • Gyeongchan Yun, Junesoo Kang, Hyunjoon Jeong, Sanghyeon Eom, Minsung Jang and Young-ri Choi, JABAS: Joint Adaptive Batching and Automatic Scaling for DNN Training on Heterogeneous GPUs, In Proceedings of the 20th European Conference on Computer Systems (EuroSys'25), 2025. [top-tier] [paper] [code] [Acceptance rate: 12.2% (85/696)]
  • Jin Yang, Heejin Yoon, Gyeongchan Yun, Sam H. Noh, and Young‐ri Choi, A Dynamic Characteristic Aware Index Structure Optimized for Real-world Datasets, ACM Transactions on Storage, (TOS), 2025. [paper]
  • Jin Yang, Heejin Yoon, Gyeongchan Yun, Sam H. Noh, and Young‐ri Choi, DyTIS: A Dynamic Dataset Targeted Index Structure Simultaneously Efficient for Search, Insert, and Scan, Proceedings of the 18th European Conference on Computer Systems (EuroSys'23), 2023. [top-tier] [paper] [code] [Acceptance rate: 16.7% (54/323)]
  • Jay H. Park, Gyeongchan Yun, Chang M. Yi, Nguyen T. Nguyen, Seungmin Lee, Jaesik Choi, Sam H. Noh, and Young‐ri Choi, HetPipe: Enabling Large DNN Training on (Whimpy) Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism, Proceedings of the 2020 USENIX Annual Technical Conference (USENIX ATC '20), 2020. [top-tier] [paper] [Acceptance rate: 18.7% (65/348)]

Talks & Presentations

  • JABAS: Joint Adaptive Batching and Automatic Scaling for DNN Training on Heterogeneous GPUs, the 20th European Conference on Computer Systems (EuroSys'25), 2025. [poster & presentation]
  • JABAS: Joint Adaptive Batching and Automatic Scaling for DNN Training on Heterogeneous GPUs, the 1st International Workshop on Systems and Methods for Sustainable Large-Scale AI (GreenSys), 2025. [workshop]