Hi, I'm Seungmin Lee
Autonomous Systems · Robotics · Computer Vision Researcher |
I build end-to-end Vision-Language-Action pipelines from data collection to distributed inference, focusing on generalization and long-horizon robotic tasks.
Autonomous Systems · Robotics · Computer Vision Researcher |
I build end-to-end Vision-Language-Action pipelines from data collection to distributed inference, focusing on generalization and long-horizon robotic tasks.
I'm Seungmin Lee, a B.S. candidate in Mechanical Engineering at Konkuk University (expected Aug 2026). I focus on autonomous systems, robotics, and computer vision, with hands-on deployment of real-time perception and control algorithms on embedded systems such as Jetson Orin Nano.
At Korea University MINT LAB, I design VLA training and data preprocessing pipelines, build real-time visualization tools for LeRobot LIBERO evaluation, and implement distributed inference with ZeroMQ for remote GPU serving. My current objective is to improve VLA generalization and long-horizon task performance for general-purpose robotics.
Korea University MINT LAB · Nov 2025 - Present
Designed VLA training scripts and data preprocessing pipelines, built real-time LeRobot LIBERO visualization, and implemented ZeroMQ-based distributed inference for remote GPU execution with closed-loop robot control.
Konkuk University · Mar 2019 - Expected Aug 2026
GPA: 3.59/4.50 · Major GPA: 3.69/4.50 · Top 9.1% · Dean's List 2025 (1st & 2nd Semester)
Dolbat · Dec 2024 - Nov 2025
Led autonomous driving team, developed perception and control algorithms, and participated in competitions.
Dolbat · Jul 2025 - Sep 2025
Secured Hanwha National Defense Award (2nd Place) by leading perception-manipulation integration with custom YOLOv8, drivable area segmentation, ONNX optimization, and ROS2 service-based robot control.
🥈 Hanwha National Defense Award (2nd Place). Built a mission-critical perception stack with custom YOLOv8 + drivable area segmentation, converted models to ONNX, and integrated ROS2 manipulator services for autonomous pick-and-place execution.
🥈 2nd Place among 26 international universities. Led perception for a 1/5-scale EV, built a custom BEV lane dataset, trained YOLOTL segmentation, and integrated ROS1 path generation with pure pursuit control.
Feel free to reach out for research collaboration in VLA, robotics, and autonomous systems.