About Me
I’m a Master’s student in Computer Science at National Taiwan University, advised by Prof. Shang-Tse Chen. I work on fairness in graph neural networks and machine learning in healthcare.
Before returning to university, I spent a year as a data scientist at a FinTech startup, building fraud- and anomaly-detection systems. My earlier training in statistics and psychology shapes how I think about measurement, behavior, and evaluation.
Outside the lab, I photograph, sketch, swim, and have trained as a lifeguard. I’ve been active in gender-equity advocacy, and I care deeply about mental health and the preservation of the Taiwanese language.
Reach me at jay.chiehen@gmail.com — I’m open to research collaboration and internships. 🤝
Current Research
- Adversarially robust fairness in graph neural networks (M.S. thesis direction): investigating how fairness-targeted attacks (e.g., node injection, adversarial edge perturbation) degrade group fairness in GNNs, and developing defenses that keep predictions both accurate and fair under attack.
- Safety and robustness evaluation of (multimodal) LLMs: building an evaluation pipeline, ALICE, which reveals what large audio-language models genuinely learn from in-context learning.
- Paternal perinatal depression: predicting depression in new fathers, testing whether maternal partner data improves prediction beyond the father’s own sleep, work–family conflict, and social support.
Education
National Cheng Kung University · Adviser: Prof. Cheng-Te Li
Thesis: Learning Graph Structures from Tabular Data for Downstream Classification Tasks
Rennes School of Business, France
National Cheng Kung University
Work Experience
National Taiwan University
Designed and graded LLM-safety labs: prompt injection, jailbreaks, hallucination detection, DPO alignment
Fazz (Singaporean FinTech startup)
Real-time anomaly detection, risk alerts, graph-based fraud analysis
NetAI Lab, National Cheng Kung University, hosted by Prof. Cheng-Te Li
Conducted Graph-ML experiments for tabular prediction. Assisted courses: Data Science, Big Data Analysis, and Social Network and Recommendation System
Health Behavior Lab, National Cheng Kung University, hosted by Prof. Carol Strong
Built the lab's R codebase for processing and analyzing longitudinal behavioral data
Selected Research
A three-stage framework that progressively reduces textual guidance to evaluate LALMs' in-context learning under audio conditioning. Built the inference framework; led results analysis and all figures.
A contrastive-learning GNN with a transformer-based extractor that jointly learns instance correlations and feature interactions for tabular prediction.
An explainable machine-learning model that predicts 14-day unplanned hospital readmission from EHR data of 24k+ patients, using SHAP-based feature analysis to align predictions with clinical experience.
Last updated on July 14, 2026