CV

Summary

Robotics and perception engineer specializing in 3D vision systems for industrial automation. My work bridges research on stereo matching, controllable depth estimation, and 6-DoF object pose estimation with deployment of robust robotic systems in manufacturing environments.

Full publication list / Patent list

Education

• Ph.D. & M.S. in Bio and Brain Engineering, KAIST, 2010-2016
• Research Trainee, Biomedical Imaging Group, EPFL, 2012-2013
• B.S. in Electrical Engineering, Hanyang University, 2005-2010

Work experience

• Mar 2024 - present: Principal Engineer
  • Samsung Electronics
  • Lead development of in-house 3D vision and robotic perception stacks, with emphasis on stereo matching, controllable depth estimation, and object pose estimation.
  • Established 3D vision systems as core sensing units for internal robotic platforms.
• Sep 2016 - Feb 2024: Senior Engineer
  • Samsung Electronics
  • Developed 3D vision algorithms and robotic perception systems for industrial automation.
  • Contributed to prototyping and deployment of box depalletizing and vision-guided automation systems.

Projects

• 3D Vision System Development (2022 - present)
  • Developed a scalable stereo matching framework, published as S²M² at ICCV 2025, with strong results across public stereo benchmarks.
  • Proposed DepthFocus, accepted to CVPR 2026, for controllable depth estimation in see-through scenes with transparent and reflective surfaces.
  • Designed and optimized stereo camera systems as perception heads for robotic automation platforms.
• Advanced Perception for Robotics (2020 - 2024)
  • Developed RGB-D based 6-DoF object detection and pose estimation models for robotic manipulation in cluttered environments.
  • Used CAD-based synthetic data and sim-to-real transfer to improve generalization across industrial components.
• Automation Solution Deployment (2018 - 2024)
  • Engineered and deployed vision-guided automation solutions, including random bin picking, depalletizing, inspection, and high-precision pick-and-place.
• Robot Learning & Manipulation (2018 - 2020)
  • Developed a 6-DoF grasp planning prototype using the Cross Entropy Method (CEM) to extend 4-DoF grasping toward cluttered scenes without relying on labeled grasp datasets.
  • Connected grasp planning with an active learning-based picking planner to improve task success rates in early industrial manipulation prototypes.
  • Implemented model-free (DDPG) and model-based (Guided Policy Search) policy learning for peg-in-hole motion generation, with a focus on contact-rich manipulation in industrial settings.
  • This early robot learning work provided useful context for later perception-driven manipulation and automation projects.

Skills

• Programming: Python, C, C++, MATLAB
• ML/CV Libraries: PyTorch, TensorFlow, OpenCV, Open3D, Scikit-learn, Pandas
• Simulation: Blender, OpenAI Gym, Mujoco, Pybullet
• Domains: 3D Vision, Robot Manipulation, Robot Learning, Factory Automation, Medical Imaging

Awards & Leadership

• 1st place, CVPR NTIRE Challenge stereo track for HR depth from specular and transparent surfaces, 2026
• 2nd place, the Samsung Creative Idea Competition, 2018
• 2nd place (out of 103), AAPM Low Dose CT Grand Challenge, 2016
• Organizer, 2nd Localization Microscopy Challenge in SMLMS, 2016
• Best Student Paper Award, IEEE ISBI, 2013
• Organizer, Localization Microscopy Challenge in IEEE ISBI, 2013
• 2nd place, Future Idea Competition, KOLON-KAIST, 2011
• National Scholarship for Ph.D. and M.S., 2010-2016