CV

Ph.D.

Korea Advanced Institute of Science and Technology (KAIST)

Artificial Intelligence

• Advisor: Prof. Jong Chul Ye (BISPL)

M.S.

Daegu Gyeongbuk Institute of Science and Technology (DGIST)

Interdisciplinary Studies of Artificial Intelligence

• Advisor: Prof. Ji-Woong Choi (Communications and Signal Processing Lab)
• Research focus: Neural decoding, bio-signal processing, cross-modal learning
• Key coursework: Advanced Deep Learning, Mathematics for Machine Learning, Brain-Machine Interface, Computational Neuroscience, Brain Disorders

B.S.

Daegu Gyeongbuk Institute of Science and Technology (DGIST)

Electrical Engineering and Computer Science

• Dean’s List (GPA 4.0 or higher out of 4.3)
• Undergraduate research: ESG-based exoskeleton robot control (UGRP)
• Key coursework: Intro to AI / Deep Learning / Computer Vision, Reinforcement Learning, Data Science, Data Structures, Algorithms, Operating Systems, System Programming, Linear Algebra, Probability & Statistics, Neuroscience I

Undergraduate Research Intern

Intelligence Augmentation Group, DGIST

• Supervisor: Prof. Jean Young Song
• Investigated Human-AI interaction methodologies and data labeling efficiency for AI training.

International Research Intern

McMaster University

• Supervisor: Prof. Qiyin Fang
• Developed smart home system frameworks for elderly care, focusing on automated sensor integration.

Student Researcher

Undergraduate Research Program (UGRP), DGIST

• Supervisor: Prof. Ji-Woong Choi
• Project: Development of an ESG-based control system for exoskeleton robots; developed a model estimating motor intention from EMG.

Software Team Member

International College Student Creative Car Competition

• Developed an object detection model.

Undergraduate Research Intern

Communications and Signal Processing Lab, DGIST

• Supervisor: Prof. Ji-Woong Choi
• Applied general linear models (GLM) to extract features from bio-signals.

OFF-CLIP: Improving Normal Detection Confidence in Radiology CLIP with Simple Off-Diagonal Term Auto-Adjustment

MICCAI 2025, early accept (top 9%)

Efficient Biomedical Time-series Contrastive Learning: Revisiting Taylor Expansion in Loss Optimization

IEEE ICTC 2025

A Study on the Trends of Adaptive Deep Brain Stimulation in Parkinson's Disease

KICS Conference 2025, oral presentation

Muscle-Guided Latent Representation Distillation for Spinal-Based Motor Intention Decoding

Lead researcher (M.S. thesis research).

• Developed an Electrospinogram (ESG)-based motor decoding framework to address the limitations of EMG/EEG-based decoding for neuroprosthetic control.
• Designed ESG-CL (contrastive learning) and ESG-DT (domain translator), which use EMG-derived latent representations to decouple motor features from motor-sensory intertwined ESG signals.
• Built the end-to-end cross-modal training pipeline, including opposite-limb and inter-subject scenarios and a simulated quadriplegia (zero-EMG) setting.

OFF-CLIP: Improving Normal Detection Performance on Chest X-ray

Co-first author, MICCAI 2025.

• Redefined the InfoNCE loss to treat all normal samples as positive pairs (off-diagonal term loss), preventing mis-clustering of semantically identical normal cases.
• Implemented all baseline models and built the benchmarking environment for comparative analysis.

Efficient Biomedical Time-series Contrastive Learning

Co-first author, ICTC 2025.

• Formulated a first-order Taylor approximation of InfoNCE that bypasses the batch-by-batch similarity matrix.
• Achieved up to 7.3x speedup and 8.5x peak-memory reduction on TS2Vec (UCI HAR) with comparable accuracy.