johnjaejunlee95's Easy AI Blog

Visitor count badge
Jae-Jun Lee profile image

AI Researcher

Jae-Jun Lee

M.S.-Ph.D. Combined Candidate @ Graduate School of Artificial Intelligence, UNIST

I study how AI models learn and generalize, with focus on meta-learning, multimodal learning, and AI4Science.

My recent interests include robustness/generalization perspectives of multimodal learning, flatness, and EEG-based AI4Science applications.

Ulsan, South Korea

Email GitHub LinkedIn Facebook Korean Blog
Meta-Learning Generalization & Robustness Multimodal Learning Robust VLA AI4Science (EEG)
3 Publications
3 Experience
5 Projects

Recent News

Latest research and updates

Selected Publications

Quick view of representative papers

  • Understanding Multimodal Learning: A Loss Landscape Smoothness Perspective
    Forty-third International Conference on Machine Learning (ICML) (2026) ·
    Paper Code
  • Can One Modality Model Synergize Training of Other Modality Models?
    The Thirteenth International Conference on Learning Representations (ICLR) (2025) · Singapore
    Paper Code
  • XB-MAML: Learning Expandable Basis Parameters for Effective Meta-Learning with Wide Task Coverage
    International Conference on Artificial Intelligence and Statistics (AISTATS) (2024) · Valencia, Spain
    Paper Code

Experience Snapshot

Research and industry collaboration

  • Machine Intelligence and Information Theory Lab (MIIT)
    Mar. 2022 - Present
    M.S.-Ph.D. Combined Research Student @ Ulsan National Institute of Science and Technology (UNIST)
    • Research in meta-learning, multimodal learning, and robust VLA.
  • Design Technology Laboratory (DTL)
    Feb. 2021 - Aug. 2021
    B.S. Internship @ Yonsei University
    • Worked on practical ML/mobile implementation projects.
  • SoC and Embedded System (SES)
    Jul. 2020 - Feb. 2021
    B.S. Internship @ Dankook University
    • Worked on embedded/SoC-related research and development tasks.

Representative Projects

Current and recent projects

AI-based Cone Beam CT Image Enhancement

Mar. 2026 - Dec. 2026

Sponsor: Dongnam Institute of Radiological & Medical Sciences

Role: Core member (MIIT, UNIST)

Development of generative model-based image enhancement for Cone Beam CT (CBCT) images.

Virtual Tactile Signal Generation Based on Multimodal Vision-Tactile AGI

Sept. 2025 - Dec. 2025

Sponsor: IITP

Role: Core member (MIIT, UNIST)

Development of 3D reconstruction containing tactile information of each segmented object.

RL Solution for High-Efficiency and Reliable Combustion System

Mar. 2022 - Aug. 2023

Sponsor: Samyang Corp. & UNIST

Role: Core member (MIIT, UNIST)

Designed a smart control system for combustion processes using deep reinforcement learning to improve efficiency and reliability.

Detailed Sections

Expand only when you need the full content

Detailed Bio
Hello! I am **@johnjaejunlee95**, a 5th-year M.S./Ph.D. integrated student at the **Graduate School of Artificial Intelligence, UNIST**. My research stems from a fundamental question: **"How do AI models learn and generalize?"** * **Past:** Meta-Learning, Generalization * **Current:** Multimodal Learning (Theoretical Approach & Applications e.g., VLA), Flatness (Loss Landscape Perspective), AI4Science (Bio) While my earlier research focused on Meta-Learning, I am currently conducting research on **Multimodal Learning**. I focus on analyzing **Multimodal Learning** through a theoretical lens, specifically connecting it to the geometry of the **Loss Landscape (Flatness)**. Recently, I have also developed an interest in **Bioinformatics (e.g., EEG)** and have begun exploring the field of **AI4Science**. My goal extends beyond merely building high-performance models; I aim to become a researcher who **"deeply understands AI based on solid theoretical and mathematical grounds and explains it clearly to the audience."** I aspire to uncover the operating principles of models both intuitively and theoretically, and to act as a bridge, conveying complex engineering concepts to the public in an intuitive and approachable way. This blog will primarily feature posts on: * **Deep Learning: Concepts, Algorithms & Intuition** * **Essential Mathematics for AI** * **Paper Reviews**
Full CV

Curriculum Vitae

Structured by section for easier updates

🎓 Education

  • Ulsan National Institute of Science and Technology (UNIST)
    M.S.-Ph.D. Combined Candidate — Graduate School of Artificial Intelligence (AIGS)
    Mar. 2022 - Present · GPA: 3.97 / 4.30
  • Dankook University
    B.S. — Department of Electronics and Electrical Engineering (EE)
    Mar. 2015 - Feb. 2022 · GPA: 3.93 / 4.50

🧪 Experience

  • Machine Intelligence and Information Theory Lab (MIIT)
    M.S.-Ph.D. Combined Research Student @ Ulsan National Institute of Science and Technology (UNIST)
    Mar. 2022 - Present
    • Research in meta-learning, multimodal learning, and robust VLA.
  • Design Technology Laboratory (DTL)
    B.S. Internship @ Yonsei University
    Feb. 2021 - Aug. 2021
    • Worked on practical ML/mobile implementation projects.
  • SoC and Embedded System (SES)
    B.S. Internship @ Dankook University
    Jul. 2020 - Feb. 2021
    • Worked on embedded/SoC-related research and development tasks.

📝 Publications

  • Understanding Multimodal Learning: A Loss Landscape Smoothness Perspective
    Jae-Jun Lee, Sung Whan Yoon
    Forty-third International Conference on Machine Learning (ICML), (2026)
    📄 Paper Code
  • Can One Modality Model Synergize Training of Other Modality Models?
    Jae-Jun Lee, Sung Whan Yoon
    The Thirteenth International Conference on Learning Representations (ICLR), Singapore (2025)
    📄 Paper Code
  • XB-MAML: Learning Expandable Basis Parameters for Effective Meta-Learning with Wide Task Coverage
    Jae-Jun Lee, Sung Whan Yoon
    International Conference on Artificial Intelligence and Statistics (AISTATS), Valencia, Spain (2024)
    📄 Paper Code

🧾 Patents

  • Single modality model-based multimodal learning device and method thereof
    Sung Whan Yoon, Jae-Jun Lee
    KR 10-2025-0066240 · Publication: May 21, 2025
  • Model-agnostic meta learning (MAML)-based domain adaptation algorithm to improve the effectiveness of developing biomedical artificial intelligence (AI) systems
    Han-Jeong Hwang, Sung Whan Yoon, Ga-Young Choi, Jae-Jun Lee
    KR 10-2024-0176258 · Publication: Dec. 02, 2024

🧩 Projects

AI-based Cone Beam CT Image Enhancement (Mar. 2026 - Dec. 2026)
Sponsor: Dongnam Institute of Radiological & Medical Sciences
Role: Core member (MIIT, UNIST)
  • Development of generative model-based image enhancement for Cone Beam CT (CBCT) images.
  • Fully contributing as a core member (MIIT, UNIST).
Virtual Tactile Signal Generation Based on Multimodal Vision-Tactile AGI (Sept. 2025 - Dec. 2025)
Sponsor: IITP
Role: Core member (MIIT, UNIST)
Partners:
- Research: Korea University Sejong Campus, UNIST, Technische Universitat Berlin
- Industry: Immertix, PSDL, SURROMIND, Sotong Five, Nongshim Data System (NDS)
  • Development of 3D reconstruction containing tactile information of each segmented object.
  • Fully contributed as a core member (MIIT, UNIST).
RL Solution for High-Efficiency and Reliable Combustion System (Mar. 2022 - Aug. 2023)
Sponsor: Samyang Corp. & UNIST
Role: Core member (MIIT, UNIST)
  • Designed a smart control system for combustion processes using deep reinforcement learning to improve efficiency and reliability.
  • Fully contributed as a core member (MIIT, UNIST).
Heterogeneous Home Appliance Data Analysis Framework (Jul. 2022 - Jan. 2023)
Sponsor: LG H&A
Role: Core member (MIIT, UNIST)
  • Designed a framework to identify causal relationships among heterogeneous smart home appliances using CCM and Meta-Learning under limited data.
  • Fully contributed as a core member (MIIT, UNIST).
Identification Data Processing Technology for On-Site Police Use (Apr. 2021 - Aug. 2021)
Sponsor: National Research Foundation of Korea (NRF)
Role: Project member (DTL, Yonsei University)
  • Developed a mobile app for facial verification for on-site use cases including missing-child identification.
  • Built with Android Studio and Java.

🧑‍🏫 Teaching & Service

📚 TA

  • Non-Parametric Bayesian Course
    Teaching Assistant · Graduate School of Artificial Intelligence, UNIST
    Sep. 2023 - Dec. 2023
  • Meta & Multi-task Learning Course
    Teaching Assistant · Graduate School of Artificial Intelligence, UNIST
    Sep. 2023 - Dec. 2023

🏫 Special Program

  • 2023 ROK Navy AI Specialist Program
    Coding sessions on fundamentals of CNN & RNN · Republic of Korea Naval Academy
    Aug. 2023 - Aug. 2023
  • LG Electronics DX Intensive
    Coding sessions on the Principles of Deep Learning · UNIST
    Summer 2024 - Summer 2024
  • LG Electronics DX Intensive
    Coding sessions on the Principles of Deep Learning · UNIST
    Summer 2022 - Summer 2022

🌱 AI Novatus Academia

  • AI Novatus Academia
    Coding sessions on the Basics of AI for Industrial Engineers · Changwon & Ulsan
    Spring 2024 - Spring 2024
  • AI Novatus Academia
    Coding sessions on the Basics of AI for Industrial Engineers · Changwon & Ulsan
    Spring 2023 - Spring 2023
  • AI Novatus Academia
    Coding sessions on the Basics of AI for Industrial Engineers · Changwon & Ulsan
    Spring 2022 - Spring 2022

🧠 Academic Experiences

  • International Conference on Learning Representations (ICLR) Reviewer
    2025, 2026
  • International Conference on Machine Learning (ICML) Reviewer
    2026

🛠️ Skills

Programming Languages
Python · MATLAB (basic understanding) · C/C++ (basic understanding)
Tools & Platforms
Linux · Bash · Git · Conda
Frameworks & Libraries
Pytorch · JAX (basic understanding) · Scikit-Learn · Pandas · Huggingface
Mathematical Backgrounds
Linear Algebra · Probability & Statistics · Information Theory (basic understanding) · Estimation Theory (basic understanding) · Learning Theory (basic understanding)
Writing Tools
Latex · Markdown · Notion
Search