의료영상과 유전체 정보를 연결하는 멀티모달 AI를 연구하고 있습니다.
작은 의료 데이터에서도 재현 가능하고 해석 가능한 모델을 만드는 데 관심이 있습니다.
I explore multimodal machine learning for biomedical research, with a focus on medical imaging, radiogenomics, graph neural networks, and reproducible experimentation for small clinical cohorts.
- Medical imaging and radiogenomics
- Heterogeneous graph neural networks
- Multimodal representation learning
- Reproducible machine-learning experiments
| Project | Contribution | Stack |
|---|---|---|
| Bayesian-GNN-Radiogenomics | MRI–genomics heterogeneous graph study for exploratory IDH mutation prediction. Public portfolio includes methodology, architecture, and aggregate results while research code and patient-level data remain private. | PyTorch, PyTorch Geometric, scikit-learn, pandas |
| KoongLog Backend | Backend contributor for an AI/IoT inter-floor-noise mediation system. Contributed APIs, analytics endpoints, data-model expansion, and integration fixes across 75 public commits. | FastAPI, SQLAlchemy, Pydantic, SQLite, Railway |
| AI-BE | AI backend contributor for a speech and Korean emotion-processing pipeline. Worked on KoBERT integration, neutral-class thresholds, file processing, and deployment fixes across 25 public commits. | FastAPI, PyTorch, Transformers, KoBERT, LangChain |
The public Bayesian-GNN-Radiogenomics repository documents the research question, heterogeneous graph architecture, experimental design, selected cross-validation results, and study limitations. Implementation and data remain private when required by research-data access, privacy, or licensing conditions.
- KCC 2026 — Encouragement Award (장려상)
- GitHub: @hozziii
- Email: nohh01089@gmail.com