A developer constantly growing by competing with yesterday's self.
Specializing in System Optimization, Diagnostic Middleware, and Efficient Deep Learning.
- Current Role: SW Engineer @ Samsung Medison (삼성메디슨)
- Education: Double Major in Computer Science & Engineering and Electronic Engineering, Kyung Hee University (Valedictorian, President's Award, GPA 4.31 / 4.5)
- Focus Areas: System Optimization, Efficient Deep Learning, Software Architecture
- Interactive Resume (CV): Open Web CV (Dark/Light & PDF Export)
- Valedictorian (President's Award, 총장상) | College of Electronics & Information, Kyung Hee Univ. (Feb 2025)
- Published Paper in J-KICS Journal (SCOPUS) | Co-first Author on adaptive multi-exit neural networks (Feb 2025)
- ICPC Asia Seoul Regional Contest | 43rd Place (Honorable Mention), National Collegiate Programming Contest Finals (Nov 2024)
- Excellence Award (2nd Place) | Kyung Hee University Programming Contest (Oct 2024)
- Best Poster Award (최우수상) | KHU Convergence Academic Festival (Dec 2024)
- National Certifications: Engineer Information Processing (정보처리기사), Big Data Analytics Engineer (빅데이터분석기사)
- Languages: OPIc IH (Intermediate High), TOEIC
- Samsung Medison (삼성메디슨) (2025.02 – Present)
- SW Engineer
- 의료 진단기기 테스트 인터페이스 및 통신 미들웨어 개발 (Development of medical diagnostic equipment test interfaces and communication middleware.)
An adaptive object detection framework dynamically balancing latency and accuracy under fluctuating computational resources.
- My Role: Co-first author; designed multi-exit backbone architecture on Faster R-CNN and formulated a dynamic exit-selection algorithm based on Lyapunov Optimization (Drift-Plus-Penalty).
- Outcome: Published in J-KICS (SCOPUS); achieved over 50% latency reduction under constrained compute resources while preserving high detection precision (mAP 70.07%).
- Tech Stack:
Python,PyTorch,Faster R-CNN,Lyapunov Optimization,PASCAL VOC / ImageNet
Leveraging intermediate model exits as specialized weak classifiers through entropy-based confidence weighting.
- My Role: Designed 10 early exits on ResNet-101 and ViT architectures; implemented temperature scaling and dynamic entropy thresholding to selectively ensemble high-confidence exit predictions.
- Outcome: Published on SSRN; improved top-1 accuracy by +1.66%p on CIFAR-100 and +0.61%p on ImageNet-1K over conventional static networks.
- Tech Stack:
PyTorch,ResNet-101,Vision Transformer (ViT),Ensemble Learning
Automated Guided Vehicle (AGV) system combining edge-level deterministic control with cloud-based fleet monitoring (SSAFY Excellence Project).
- My Role: System architect and lead developer; built C++ AGV control daemon on Raspberry Pi and MQTT/WebSocket streaming pipeline connected to AWS EC2 and MySQL.
- Tech Stack:
C++,Python,Raspberry Pi,AWS EC2,MQTT,MySQL
Automated university portal grade tracker and instant notification extension used by fellow students.
- My Role: Implemented secure session handling, periodic background polling with Chrome Alarms/Storage APIs, and regular expression parsing of university intranet tables.
- Tech Stack:
JavaScript,Chrome Extension API,HTML/CSS
