I am a Ph.D. Student at KAIST AI, advised by Prof. Han-Lim Choi in the Laboratory for Information and Control Systems (LiCS). My research interests include vision-language models, multi-agent systems, and intelligent robotics. I am particularly interested in how robot swarms coordinate and make decentralized decisions under limited sensing, communication, and computation. My long-term interests also include superintelligence and how learning and reasoning can enable more capable autonomous systems.
* Equal contribution
Asian Conference on Computer Vision (ACCV)
Decoding local geometry from frozen CLIP tokens through fixed neighborhood aggregation, gated residual updates, and shallow token-detail refinement.
IEEE International Conference on Robotics and Automation (ICRA)
Reusing internal register tokens to provide spatial context for action generation without increasing the model's parameter count.
IEEE Access
Connecting multimodal perception and planning for collaborative agent systems.
Engineering Applications of Artificial Intelligence
Exploring model merging to combine the capabilities of Korean and multilingual language models.
Asia-Pacific International Symposium on Aerospace Technology (APISAT)
Learning spatial embeddings from monocular images to improve imitation learning for robotic manipulation in spacecraft docking assistance.
Journal of the Korean Society for Aeronautical and Space Sciences
Reviewing airworthiness policies, certification standards, and safety assurance approaches for AI-based military aircraft software, with a focus on autonomous flight.
Preprint
Injecting depth-derived geometry features into visual tokens through a learned gated residual connection for vision-language-action models.
Preprint
Fusing monocular depth with multi-view RGB observations to improve a diffusion policy's tolerance to camera loss in simulated spacecraft berthing.
Preprint
Compact zero-shot anomaly detection evaluated across industrial benchmarks and deployed on real edge devices.
Preprint
Reading geometric cues from a frozen CLIP vision encoder with a compact decoder and no text encoder at inference.
I founded Plaid Labs, where we help businesses modernize their operations through Digital Transformation (DX) and AI Transformation (AX). We build and integrate software and AI systems, from workflow automation and private LLM deployment to edge and on-device AI connected to existing hardware.
온디바이스 결함 탐지 장치 및 이를 활용한 결함 탐지 방법
