About Me

I'm a first-year M.S./Ph.D. integrated student at KAIST Graduate School of AI (GSAI), supervised by Prof. Jaegul Choo. My research interests include modeling lifelike avatars — particularly relightable and animatable representations — and subject-conditioned video generation.

I received my B.E. in Data Science from Hanyang University. While pursuing the degree, I worked as an AI researcher on 2D face swapping at dob Studio Inc., and as a research intern at the Samsung Electronics AI Center.

Experience

KAIST
Samsung
DePaul
dob Studio
Hanyang
View details Hide details
  • Sep 2024 – Feb 2026
    KAIST AI
    Research Intern, DAVIAN Lab.
  • Mar 2025 – Jun 2025
    Samsung Electronics
    Research Intern, AI Center.
  • Nov 2022 – Sep 2024
    DePaul University
    Research Intern, Data Science Innovation Lab.
  • Sep 2023 – Oct 2024
    dob Studio Inc.
    AI Researcher, R&D Center.
  • Dec 2023 – Feb 2024
    Seoul National University
    Explore CSR Student, SKI-ML Lab.
  • Oct 2020 – Feb 2021
    Hanyang University
    Research Intern, Data Science Innovation Lab.

Publications

* denotes equal contribution

2026

  • EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure

    *Junyeong Ahn, *Seojin Yoon, Sungyong Baik

    arXiv preprint

    Paper

  • TexAvatars: Hybrid Texel-3D Representations for Stable Rigging of Photorealistic Gaussian Head Avatars

    *Jaeseong Lee, *Junyeong Ahn, Taewoong Kang, Jaegul Choo

    3DV 2026

    Paper Project Page
2024

  • Digital Health Sensor Data in Autism: Developing Few Shot Learning Approaches for Traditional Machine Learning Classifiers

    Casey C Bennett, Cedomir Stanojevic, Jiyeong Oh, Junyeong Ahn, Yeeun Jeon, Kanghee Son, Nikki Abbott

    IEEE BSN 2024

    Paper

Projects

2023

  • Identity-Preserving Face Swapping

    dob engine 2.0 · dob Studio Inc.

    • Built CrossSwap, a face swap model that preserves source identity features via cross-attention between source and target feature maps, boosting the identity preservation metric by 10% over SOTA.
    • Developed a shape-aware identity extractor combining identity embedding with 3DMM to retain source facial shape.