M.S. Researcher in AI Convergence Engineering @ Sungkyunkwan University
Trustworthy AI · Computer Vision · Multimodal Learning · Graph Mining
I study how AI systems can make reliable decisions grounded in verifiable evidence. My work spans medical imaging, anomaly detection, multimodal reasoning, and large-scale data analysis—from designing research questions to implementing models and evaluating their robustness.
ACCV 2026 · Accepted
- Studied why high predictive confidence does not necessarily imply reliable visual evidence in chest X-rays.
- Developed an evidence-aware selective prediction framework and evaluated evidence mismatch using perturbation-based analyses.
- Role: Led research design, implementation, experiments, analysis, and paper preparation.
PAKDD 2026 · Oral presentation
- Contributed to a dataset for anomaly detection on reflective pouch-type battery surfaces, including spatial zone annotations.
- Focus: Industrial computer vision and anomaly detection.
Ongoing research
- Studying distribution shift, calibration, and selective prediction using clinical time-series data.
- Focus: Evaluation design, robustness, and risk-aware decision support.
Collaborative research project
- Analyzed approximately 1.56 million text–GIF interactions.
- Investigated semantic communities, graph structure, and the relationship between semantic diversity and reuse.
- Conducted statistical validation and robustness checks.
| Year | Work | Venue / Status |
|---|---|---|
| 2026 | Confidence Is Not Evidence: TAESAN for Evidence-Aware Selective Prediction in Chest X-Rays | ACCV 2026 · Accepted |
| 2026 | BatterySurAD: A Dataset for Anomaly Detection on Pouch-Type Reflective Battery Surfaces with Spatial Zone Annotations | PAKDD 2026 · Oral |
| 2026 | AIMER: Affective Intention-guided Multimodal Emotion Reasoner for Visual Emotion Analysis in Social Media | WWW 2026 · Oral · Co-author |
| 2026 | More Modality, Less Reliability? A Negative Result on MLLM-Generated Clinical Knowledge Claims | KDD 2026 · Poster |
| 2025 | MAMBA-PAD: Selective State Space Anomaly Detection with Channel-wise Gating and Stabilized Dual-Objective Training | ICDM OWAD Track · Poster |
For a complete publication list, see Google Scholar.
Programming & data
Research methods
Computer Vision · Anomaly Detection · Multimodal AI · Graph Mining · Statistical Analysis · Calibration · Selective Prediction
Additional tools
SPSS · Figma · Excel
- 1st Place, AI Challenge for Biodiversity — SK hynix (2022)
- Excellence Award, Shinhan Financial Group Big Data Hackathon, Shinhan Card Track (2022)
- Excellence Award, Digital Power On — Hana Financial Group (2023)
📌 Code, reproducibility materials, and project links will be added as they become available for public release.
📚 Google Scholar · 🔗 LinkedIn is linked in my GitHub profile.