I am currently conducting mental health research at Dr. Presso. Our first project—carried out in collaboration with professors from Seoul National University and Korea University College of Medicine—focused on developing an AI model capable of diagnosing PTSD four years in advance, based on initial patient evaluations (questionnaires) and blood test data.We are now gathering both Active and Passive EMA data on depression, aiming to advance our understanding and management of depressive disorders. To achieve this, we’re building a counseling AI powered by an RAG-based large language model (LLM). Ultimately, our goal is to develop an accessible service that helps people easily detect and treat depression.

Prostate Cancer Diagnostic AI Algorithm Development (Mar 2023 – Apr 2024)[Goal]- Develop a clinical + genomic AI model to reduce unnecessary prostate biopsies (Target AUC ≥ 0.74).[Achievements]- AI Performance: Achieved AUC 0.93 by integrating EMR and genomic data.Patent: Filed a patent (two-stage learning-based prediction system, Nov 2023).[Publications]- Co-author on Journal of Urologic Oncology (DOI: 10.22465/juo.244800400020)- First author paper under revision for BMC Medical Informatics and Decision Making- Validation: KCL-certified performance verification and retrospective clinical trial.[Key Responsibilities]Integrated AI Model Development:- Combined EMR-based AI and genomic risk models, achieving a final AUC of 0.93EMR-based AI Model:- Conducted feature selection, EDA, and data preprocessing- Tested multiple ML models, optimized hyperparameters, and performed model stacking- Identified key risk factors influencing prostate cancerGenomic Risk Model:- Processed and analyzed microarray data using DNNs- Addressed overfitting via GWAS data, achieving AUC 0.661 with genomic data alone- Outperformed a Nature 2023 study’s genomic-only model (AUC 0.627)Tech Stack: Python (PyTorch, Scikit-learn), PLINK; leveraged SNP-based genomic data and EMRs.