Seongsu Kim

Pricinple Investigator: Sungsoo Ahn
Language: Korean, English
Email: seongsu.kim@kaist.ac.kr
Hi, I’m Seongsu. Thank you for stopping by! 👋 The spelling of my name in Korean is “성수” and it is pronounced as [sʌŋ-su:].
I am currently an incoming Ph.D. student in the AI department at KAIST. I am also a member of Structured and Probabilistic Machine Learning (SPML) Lab. I completed my M.S. in Artificial Intelligence at POSTECH.
My research interests include integrating AI into scientific research and using AI to uncover scientific facts. Also, I am interested in physical and chemical concepts like solid state physics, simulation dynamics and molecular science, and also mathematical concepts like group theory, geometry and geometrical deep learning.
Recently, I am focusing on (1) generative modeling for materials and (2) ab-initio method based machine learning for quantum physics and chemistry related with density functional theroy (DFT) or variational Monte Carlo (VMC).
I believe that AI will effectively reduce the complexity of calculations in physics and chemistry, especially in areas like multi-objective optimization problems, challenges in establishing ansatzes, modeling interactions and governing equations, and handling intractable computations. I am studying and researching these possibilities.
I am always open to discussion, so please feel free to reach out anytime!
News
Sep 18, 2025 | 📋 Two papers were accepted to NeurIPS 2025: one Spotlight (first author) and one Poster (second author). |
---|---|
Jun 20, 2025 | 🏫 I will be starting my Ph.D. program at KAIST under the supervision of Sungsoo Ahn |
May 25, 2025 | 🐋 Two paper, “High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction” and “Flexible MOF Generation with Torsion-Aware Flow Matching” are updated. |
May 20, 2025 | 📕 Honorably, I was selected as Notable Reviewer at 2025 ICLR! |
Jan 23, 2025 | 💎 One paper, “MOFFlow: Flow Matching for Structure Prediction of Metal-Organic Frameworks”, has been accepted to 2025 ICLR. |
Oct 31, 2024 | 💎 One paper, “MOFFlow: Flow Matching for Structure Prediction of Metal-Organic Frameworks”, has been accepted to AIDrugX workshop at 2024 NeurIPS. |
May 16, 2024 | 🚀 One paper, “Gaussian Plane-Wave Neural Operator For Electron Density Estimation”, has been accepted to 2024 ICML. |
Publications
- NeurIPSPosterFlexible MOF Generation with Torsion-Aware Flow MatchingNeural Information Processing Systems (NeurIPS) 2025
- ICLRPosterMOFFlow: Flow Matching for Structure Prediction of Metal-Organic FrameworksInternational Conference on Learning Representations (ICLR) 2025