HI, I'm JOON HYUK LEE 👋
I've worked both sides of a problem: analyzing data to find where the real gap is, and building the thing that closes it. What I took from it is that a model disconnected from a real problem stays a model.
So the part I care about is the link between them — reading the context of an industry, following where the data points, and putting it in language the people who decide can act on. More in my Résumé.
안녕하세요, 이준혁입니다 👋
데이터를 분석해 진짜 공백이 어디인지 찾고, 그 공백을 메우는 결과물을 직접 만드는 일을 좋아합니다. 그 과정에서 얻은 확신은, 현장의 문제와 연결되지 않은 모델은 결국 모델로만 남는다는 것이었습니다.
그래서 제가 관심을 두는 지점은 둘 사이의 연결입니다. 산업의 맥락을 읽고, 데이터가 가리키는 곳을 따라가고, 그것을 의사결정하는 사람이 실행할 수 있는 언어로 옮기는 일입니다. 자세한 내용은 제 Résumé에서 확인하실 수 있습니다.
Research Interests
My Skills
My Latest Work
2026
Paper SubmissionMIDL 2026A Comparative Study of Deep Learning-Based Anomaly Detection for Intracranial Hemorrhage in Brain CT
A systematic benchmark of 11 deep learning anomaly detection methods for intracranial hemorrhage detection in brain CT favors reconstruction-based over self-supervised approaches.
2026
ProjectMusinsa SKIN-Fit — Ingredient-Based Skincare Recommendation
An ingredient-based skincare recommendation service for Musinsa Beauty. From ~700K reviews and 1,198 products, it shows the platform doesn't use the skin-type data it collects, then closes the gap with an ingredient-OCR → skin-fit scoring pipeline surfaced as a Chrome extension.
2025
Paper SubmissionKSC 2025Multimodal Biosignal-Based Attention Prediction Model
Proposes a machine learning pipeline to predict learner attention using multimodal wearable biosignals — HR, RRI, EDA, and skin temperature / beyond single-sensor approaches by leveraging multimodal time-series data in educational attention monitoring.
B.S. in Software Convergence · B.S. in Physical Education
- GPA 3.91 / 4.5
- KHUDA 경희대학교 데이터분석 & AI 학회
- AIMS Lab 학부연구생 (2025.08 - 2026.07)