Research
I'm an AI researcher with a civil engineering background. My work spans 3D perception, agentic AI, and AI safety, and I care most about research that leaves the lab and holds up in the real world. I like to build end to end, from field sensing systems to the learning algorithms behind them.
Research Directions
013D (Gaussian Splatting, LiDAR)
Real-time 3D reconstruction from LiDAR and cameras: LiDAR-inertial odometry, mobile mapping, and neural representations such as NeRF and 3D Gaussian Splatting. I shipped these ideas as DepthViz, an on-device iOS LiDAR-SLAM scanner, and study lightweight odometry and open-vocabulary 3D scene understanding for robots.
02Agentic AI
Agents that act reliably in the real world: mobile GUI agents, multi-agent orchestration, and on-device collaborative behavior. My work on MATE (Mobile Agent Trustworthy Execution) makes agent actions safe and efficient, and I bring the same ideas to ontology-based home robots and multi-user memory control.
03AI Safety
Making AI systems trustworthy: safety evaluation for multimodal generative AI, robust detection of AI-generated images and video, and efficient, verifiable inference. This line runs from agent-level safety to content-level trust, connecting my work on generative-content detection and reliable execution.
Real-World Problems
More than any single method, this is the area I care about most: research that leaves the lab and solves an actual problem in the field.
Detection-driven super-resolution for sonar imagery (IEEE TGRS) for mine-like object detection, UAV SAR and LiDAR digital twins for landslide and disaster mapping, and remote sensing for environmental monitoring. I want work where the benchmark is the real world, not just a dataset.
Funded Research Projects
- ETRI — Multimodal AI-based Digital Twin Model Automatic Generation (Apr. 2026 – Present)
- NCSL — Safety Evaluation Technology for Multimodal Generative AI (Mar. 2026 – Present)
- IITP — On-Device AI Inter-Object Collaborative Behavior (Apr. 2025 – Present)
- LIG Nex1 — Underwater Sonar Image Enhancement (Apr. 2025 – Jul. 2026)
- SOOMVI — Precise Landing Technology for Drones (Jun. 2025 – Apr. 2026)
- Stellarvision — Sensor-Fusion (SAR, LiDAR) UAV Digital Twin Platform for Landslides (Jul. 2024 – Feb. 2025)
See the full Publications list for papers.