Minguk Choi

mgchoi@dankook.ac.kr
Seoul, SouthKorea
I will begin my Ph.D. in Computer Sciences at the University of WisconsinโMadison in Fall 2025.
Currently, I am conducting research under the supervision of Prof. Matthias Boehm at TU Berlin, while also collaborating with Dr. Kyoungmin Kim at EPFL.
Previously, I earned my Masterโs degree in August 2024 from Dankook University in Korea, where I had the privilege of being advised by Professors Seehwan Yoo and Jongmoo Choi in the System Software Laboratory.
My research focuses on scalable and efficient systems for ML training, retrieval, and serving. Here is my CV.
On-going projects:
-
Federated Learning Plan under Privacy Constraints: Optimized federated execution plans for end-to-end ML pipelines in Apache SystemDS under privacy constraints, using a dynamic programming-based cost model.
-
Hybrid Vector-Relational Search: An algorithm to optimize multi-vector top-k queries combined with complex relational filters in vector-relational hybrid databases.
Last updated: April 6, 2025
news
Mar 14, 2025 | Honored to receive the SIGMOD 2024 Best Artifact Awardโlooking forward to receiving it at SIGMOD 2025. See you in Berlin! |
---|---|
Aug 31, 2024 | I am starting research on Apache SystemDS, remotely supervised by Professor Matthias Boehm! |
Aug 22, 2024 | I graduated with a Masterโs degree from Dankook University! |
Jun 17, 2024 | At Korea Computer Congress 2024 in Jeju, Korea, five of our papers were accepted, and I presented our SIGMOD paper. We were honored to receive a Certificate of Appreciation, the Best Paper Award, and the Best Presentation Award. |
Jan 22, 2024 | Our paper โCan Learned Indexes be Built Efficiently? A Deep Dive into Sampling Trade-offsโ has been accepted with minor-revision for SIGMOD โ24 (Round 4)! |
latest posts
selected publications
- KCCAnalysis of RMI Using CPU-Optimized Search AlgorithmsKorea Computer Congress, 2024
- KCCBreakdown Internal Operations in Updatable Learned IndexKorea Computer Congress, 2024