Finance & Documents | Industrial Equipment | Model Efficiency
I turn AI research into systems that work in the field.
- 🏦 Finance & Documents: Handwritten tax-bill automation (40 doc types, 3 s/doc, 96% accuracy), efficient document VLMs
- 🏭 Industrial Equipment: Anomalous sound detection, vibration fault diagnosis, MCP server for equipment monitoring
- ⚡ Model Efficiency: 83% fewer parameters in multi-task learning (IEEE Access, 1st author)
- 📫 popo2419@naver.com
- Efficient and adaptive AI systems: Multi-task learning, dynamic model optimization, pruning, and parameter-efficient inference
- Multimodal representation learning: Vision-language models, document understanding, and token/parameter reduction
- Industrial AI and signal intelligence: PHM, vibration-based fault diagnosis, time-series modeling, and robustness under noise
| Year | Paper | Venue | Affiliation | Links |
|---|---|---|---|---|
| 2026 | Physics-R: Reliability-Gated Closed-Loop Acquisition of Multimodal Evidence for Fault Diagnosis Under Unseen Operating Conditions Choi S., Shin Y. |
IEEE Access · under review | Korea Electronics Technology Institute |  |
| 2026 | Asymmetric Memory Retrieval for Training-Free First-Shot Anomalous Sound Detection Choi S., Yang S., Shin Y. |
IEEE Signal Processing Letters · under review | Korea Electronics Technology Institute | |
| 2026 | Index-Preserving Lightweight Token Pruning for Efficient Document Understanding in Vision-Language Models Son J., Choi S., Yun I. |
ICLR 2026 Workshop on MM Intelligence · Poster | Hana Institute of Technology |    |
| 2023 | Task-Aware Dynamic Model Optimization for Multi-Task Learning Choi S., Jin H., Kim E. |
IEEE Access | Chung-Ang University |  |
