Youngwoong Cho

I'm an M.S. Robotics student at Georgia Tech, advised by Prof. Danfei Xu in the Robot Learning and Reasoning Lab (RL2). I work on generalizable robot policies: from one dexterous hand to another, and from human demonstrations to robots.

Previously, I was a Research Engineer at RLWRLD, adapting robot foundation models to dexterous hands. Before that, I interned at Naver Labs and Seoul Robotics, and received my B.E. in Mechanical Engineering from The Cooper Union.

I'm looking for research internships for Summer 2027. Feel free to reach out!

Email / CV / GitHub / LinkedIn

Youngwoong Cho

Timeline

  1. Aug 2026 – presentGeorgia Tech RL2Graduate Researcher, advised by Prof. Danfei Xu
  2. Apr – Aug 2026RLWRLDResearch Engineer
  3. Mar – Aug 2022Naver LabsDeep Learning Research Intern, Autonomous Driving
  4. May 2021 – Mar 2022Seoul RoboticsMachine Learning Research Engineer Intern
  5. May – Jul 2020Seoul National UniversityResearch Intern, Prof. Dongjun Lee

Research

One Action Interface, Seven Dexterous Hands
Georgia Tech RL2, advised by Prof. Danfei Xu
[Project page]

Can a policy trained on several robot hands control a hand it has never seen?

Cross-embodiment · Dexterous hands · Isaac Sim · Vision Pro teleoperation

Vision Pro Teleoperation for Dexterous Hands
Georgia Tech RL2
[Project page]

I built an Apple Vision Pro teleoperation setup for seven simulated robot hands and recorded 357 grasping demonstrations with it. On hardware, I teleoperate a Wuji Hand 2 with the Vision Pro to collect tactile data.

Apple Vision Pro · Teleoperation · Isaac Lab · Tactile sensing

Humanoid Manipulation from Human Demonstrations Only
Georgia Tech RL2, part of the WARP project
[Project page]

Policies trained only on egocentric human demonstrations, with no robot data, run zero-shot on a real RB-Y1 humanoid. Along the way I studied what makes retargeted demonstrations learnable.

Human demonstrations · Point clouds · Zero-shot transfer · Rainbow RB-Y1

Adapting Robot Foundation Models to a 16-DoF Hand
RLWRLD, Research Engineer
[Project page]

At RLWRLD I worked on getting large pretrained policies (GR00T N1.6, π0.5, and a 3D geometry-aware policy) to work on DexJoCo, a public benchmark with 11 dexterous manipulation tasks.

GR00T N1.6 · π0.5 · GAM · DexJoCo

Splitting Robot State into Tokens in GR00T N1.6
RLWRLD, Research Engineer
[Project page]

GR00T N1.6 packs a humanoid's 48 joint values into a single token. Splitting them into 11 tokens raised simulated success from 35.8% to 41.5% without adding any parameters.

GR00T N1.6 · VLA · Humanoid · Isaac Sim

Projects

BarlowZero: Sample-Efficient Model-Based Reinforcement Learning
[Report] | [Code]
ECE-472 Deep Learning, The Cooper Union · Fall 2022

A MuZero variant with Barlow Twins self-supervision. It cut the training batch size from 512 to 64 while performing comparably to EfficientZero on Atari Breakout.

Self-Driving Delivery Robot
Senior capstone, The Cooper Union · Spring 2023

An end-to-end autonomous navigation stack (3D LiDAR SLAM, localization, planning, control) for an indoor delivery robot, with localization latency under 100 ms.

Cooper Union IGVC
Intelligent Ground Vehicle Competition · 2022–2023

LiDAR- and RGB-based localization, mapping, and path planning in ROS, plus an OCR-based stop-sign classifier. Placed 2nd in autonomous navigation and 3rd in vehicle design.

Education

  • Georgia Institute of Technology
    M.S. in Robotics
    Expected May 2028
  • The Cooper Union
    B.E. in Mechanical Engineering, Minor in Computer Science, Summa Cum Laude
    2023