TwinVLA: Data-Efficient Bimanual Policy through Twin Single-Arm Vision-Language-Action Models
Preprint, 2025
Project Page · Paper · Code
Euijin Jeong
AI Researcher
I work on what I think is cool @ RLLAB Yonsei, advised by Professor Youngwoon Lee. My current interest is compression, continual learning, and meta-RL.
Publications
Experience
Intern - RLLAB @ Yonsei Univ.
Seoul, South Korea, Nov 2024 - Present
ML Engineer - FAiKERZ.Inc
Seoul, South Korea, Oct 2023 - Dec 2024
- Developed and deployed AI solutions for detecting copyright infringement
- Developed an e-commerce item clustering algorithm
- Built MLOps for on-premise continuous deployment
RL Research Scientist - Medipixel.Inc
Seoul, South Korea, Apr 2020 - Aug 2021
- Built RL agents and pipelines for automated cardiovascular stent procedures for animal experiments
- Developed Offline RL using clinical trial data
- Created a 3D sim-to-real cardiovascular simulation for zero-shot transfer RL
Undergraduate Researcher - CIIAB @ GIST
Gwang-ju, South Korea, Mar 2019 - Dec 2019
- Built Imitation learning agents for StarCraft II
- Developed RL algorithms (TD3, PPO)
Projects
hcam-torch
Feb 2023
- Implemented Multi-modal LSTM-PPO of Hierarchical Memory Reinforcement Learning paper (2105.14039)
- Created an OpenAI-Gym wrapper for an RL reasoning environment introduced in the paper
simple-es & pyNeat
Sep 2021
- Implemented multi-agent evolution and genetic algorithms (ES, GS, OpenAIES) in the PettingZoo environment
- Developed a distributed learning NEAT algorithm using mpi4py
rl-algorithms
Jul 2021
- Maintained open-source algorithms developed at Medipixel.Inc
- Implemented an R2D1 (single-actor R2D2) agent
kerbal-rl & ksp-ai
Nov 2019
- Developed an RL environment for the rocket simulation game KSP
- Trained a rocket hovering control agent using SAC