Seungwon (SW) Choi
PhD Candidate at UCLA and 2025 Amazon Fellow
Developing transformer-based networks for robot learning using multimodal egocentric data from head-mounted devices.
Intro
I am Seungwon (SW) Choi, a PhD Candidate at UCLA and a 2025 Amazon Fellow specializing in robotics and AI. My work sits at the intersection of mechanical engineering and machine learning, where I focus on 'expert embodiment'—the process of transferring implicit human skills to robots through imitation learning and multimodal egocentric data. Throughout my career, from medical robotics research in Korea to my current work at UCLA, I have been passionate about building robots that can safely and intelligently navigate the real world, whether for infrastructure inspection or surgical assistance. I am always grateful for the opportunity to connect with fellow researchers and engineers who are advancing the frontiers of human-centric AI and robotics.
Networking
What I can offer
- ›Expertise in training robot policies using human-centric multimodal data
- ›Experience in building real-world robotic hardware for inspection and medical use
- ›Technical proficiency in NVIDIA Isaac Sim and real-time control
What I'm looking for
- ›fellow researchers in the AI/Robotics space
- ›the Amazon Fellow cohort
- ›expanding my professional network
Best fit for
Focus
Current interests
Core competencies
Background
Career
Progressed from research internships in medical robotics and motion control to a Master's and PhD focused on robotics, culminating in a prestigious Amazon Fellowship and research roles at UCLA.
Education
Ph.D. in Mechanical and Aerospace Engineering (Robotics), UCLA (2021 – 2026 expected); M.S. in Mechanical Engineering (Robotics), Korea University (2018 – 2020); B.S. in Mechanical Engineering, Korea University (2011 – 2018)
Achievements
- ›Selected as 2025 Amazon Fellow (1 of 13 fellows)
- ›Approved Form I-140 (EB-2 NIW) in March 2025
- ›Presented at IEEE ICRA 2024 and NASA JPL Robotics Seminar
- ›Built and operated complex mobile manipulators integrating Neuromeka and AgileX hardware
Opinions
- Robots should learn from 'expert embodiment' by transferring implicit human skills to robotic policies.
- Mechanical Engineering must be bridged with AI, Computer Science, and AR/VR for true robotic advancement.
Personality
Communication style
Professional, academic, and highly appreciative with a focus on technical precision.
Formality — 8/10
Vocabulary