Minchan Jeong
Ph.D. in AI (expected Aug 2026), Kim Jaechul Graduate School of AI, KAIST
Advisor: Se-Young Yun
· Expected graduation: 2026-08
I'm finishing my PhD at KAIST, advised by Se-Young Yun (Thesis: Operator Spectral Estimation Beyond IID Data). I build neural methods that recover an operator's leading spectrum with provable guarantees, for analyzing stochastic dynamical systems and for computing the excited states of quantum many-body systems. Two results anchor this: a parametric SVD of the Koopman operator (NeurIPS 2025) and low-rank variational Monte Carlo for quantum excited states (NestedLoRA-VMC), both with Jon Ryu.
Earlier I led a mid-term precipitation post-processing ML model on a terabyte-scale data pipeline (2022 – Spring 2023), then built the JAX inference engine behind an operational GraphCast deployment at KMA (Fall 2023 – 2024). I'm drawn to problems where the mathematics and the systems both have to be right.
Education
- PhD in AI, KAIST
- BS in Physics and Mathematics, Seoul National University
Selected publications
- Variational Monte Carlo for Quantum Excited States via Nested Low-Rank Approximation2026 . Manuscript under review, 2026.quantumspectral
- 2026 . To appear in Transactions of the Association for Computational Linguistics (TACL).llm
- Advances in Neural Information Processing Systems, 2025, pp. 25564–25600spectral
- Advances in Neural Information Processing Systems, 2022, pp. 4231–4243spectral
- Advances in Neural Information Processing Systems, 2022, pp. 38461–38474federated learning
Recent
- — New manuscript on variational Monte Carlo for quantum excited states.
- — Group-Normalized IVO accepted at ICLR 2026. (coauthor)
- — Koopman SVD presented at NeurIPS 2025.