cv
General Information
| Full Name | Dongwon Son |
| dongwon.son110@gmail.com | |
| Website | dongwon-son.github.io |
| Research Interests | Robot learning for manipulation; computationally efficient simulation; planning; sim-to-real transfer |
Education
- 2022 - Present
PhD Student, AI Graduate School
Korea Advanced Institute of Science and Technology (KAIST), Seoul, Korea
- {"Advisor"=>"Beomjoon Kim"}
- 2018 - 2020
Master's Degree, Mechanical Engineering
Seoul National University, Seoul, Korea
- {"Advisor"=>"Dongjun Lee"}
- Thesis: Multi-Contact Simulator and Reinforcement Learning for Screw Tightening Tasks
- 2010 - 2014
Bachelor's Degree, Mechanical Engineering
Seoul National University, Seoul, Korea
Experience
- 2014 - 2017
QA Engineer, Gas Turbine Engine Division
Samsung/Hanwha Techwin, Changwon, Korea
- 2020 - 2022
AI Researcher, AI Methods Team
Samsung Research, Seoul, Korea
- Developed robot learning algorithms for vision-based manipulation, including cluttered-scene grasping, sim-to-real transfer, planning, and data-efficient reinforcement learning.
- 2025.10 - 2026.6
Research Intern, Robotics Group
Allen Institute for AI (Ai2), Seattle, WA
- Worked on a torque adaptation module to reduce the sim-to-real gap under the mentorship of Dieter Fox.
Skills
| Languages | Python, C++, MATLAB |
| Frameworks | TensorFlow, Keras, JAX, ROS, OpenGL, PyBullet, Open3D, OMPL, FCL |
| Tools | SolidWorks, EasyEDA |
Selected Research Topics
- Robot manipulation, including object manipulation, grasping, vision-based assembly, and long-horizon manipulation
- Computationally efficient simulation, including physics engines, collision detection, and rendering
- Sim-to-real transfer for manipulation and assembly tasks
- Structured network design and shape representation for planning
- Open-source, cost-efficient manipulator design