CV
Kiyoung Om
Research Intern at NAVER LABS; KAIST M.S.
Summary
Research intern at NAVER LABS and M.S. graduate from KAIST Graduate School of Data Science. My current research focuses on closed-loop covariate shift in traffic simulation for autonomous driving, with broader interests in generative models for optimization, control, and alignment.
Education
- M.S. in Data ScienceFeb. 2026Korea Advanced Institute of Science and Technology (KAIST)Courses: Graduate School of Data Science, Advisor: Prof. Jinkyoo Park, System Intelligence Lab
- B.S. in Industrial Management EngineeringFeb. 2024Korea UniversityGPA: 4.02/4.5
Work Experience
- Research InternMar. 2026 - Aug. 2026NAVER LABSResearch on closed-loop covariate shift in traffic simulation for autonomous driving.
- Studying robust simulation and learning methods for autonomous vehicle policies interacting with dynamic traffic environments.
- Focusing on distribution shift induced by closed-loop policy-environment interaction.
- AI Research AssistantMar. 2025 - Feb. 2026Samsung Electronics-KAIST AI Industry-Academia CollaborationAI-enhanced supply chain management optimization.
- Developed subgradient lambda prediction methods for complex SCM optimization.
- Improved optimization speed and efficiency through machine learning approaches.
Skills
Research
- Autonomous driving
- Traffic simulation
- Closed-loop covariate shift
- Sequential decision making
- Black-box optimization
- Generative model alignment
Methods
- Diffusion models
- Flow-based models
- Reinforcement learning
- Bayesian optimization
- Posterior inference
Publications
- Diffusion Fine-Tuning via Reparameterized Policy Gradient of the Soft Q-Function2026
- Diffusion Alignment as Variational Expectation-Maximization2026ICLR 2026Jaewoo Lee, Minsu Kim, Sanghyeok Choi, Inhyuck Song, Sujin Yun, Hyeongyu Kang, Woocheol Shin, Taeyoung Yun, Kiyoung Om, and Jinkyoo Park.
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization2025NeurIPS SPIGM Workshop 2025 (Oral)Kiyoung Om, Kyuil Sim, Taeyoung Yun, Hyeongyu Kang, and Jinkyoo Park.
- Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization2025ICML 2025; FPI @ ICLR Workshop 2025Taeyoung Yun, Kiyoung Om, Jaewoo Lee, Sujin Yun, and Jinkyoo Park.
Presentations
- Geometry-aware Posterior Inference for High-dimensional Black-box Optimization2025KAIST SILAB SeminarDaejeon, South KoreaSeminar on generative model-based optimization on Riemannian manifolds.
- Control as Probabilistic Inference2025KAIST SILAB SeminarDaejeon, South KoreaSeminar on interpreting control problems through probabilistic inference.
Teaching
- IE343 Statistical Machine Learning2024KAIST, Industrial & Systems EngineeringRole: Teaching AssistantGraded assignments, provided feedback, and held office hours.
- Deep Learning Programming Study2023Korea UniversityRole: Teaching AssistantHelped participants implement deep learning pipelines with NumPy, PyTorch, AutoGrad, and PyTorch-Lightning.
Languages
- KoreanNative
- EnglishFluent
Interests
- Autonomous driving and simulationTraffic simulation, Closed-loop evaluation, Distribution shift
- Generative decision makingOptimization, Control, Alignment