Machine Learning
Developing robust and principled models that connect theory with real-world healthcare and NLP applications.
Ph.D. Student · Computational Biology and Bioinformatics · Duke University
I study how models remain reliable under uncertainty.
I am a Ph.D. student at Duke University studying reliable machine learning under uncertainty. My work draws on statistics and causal inference to understand incomplete, biased, and evolving data in healthcare and other high-stakes settings.
I study reliable machine learning under incomplete, biased, and heterogeneous data.
I am interested in missing data, distribution shift, and multimodal learning, and in how these factors affect model reliability and downstream decisions. I draw on tools from causal inference, representation learning, and modern deep learning, with a focus on healthcare and other high-stakes domains.
- Multimodal Learning under Missingness: modeling MNAR patterns and informative missing signals in clinical and recommendation settings.
- Robust Generalization and Domain Adaptation: building adaptive frameworks that improve cross-dataset transfer and calibration.
- Causal and Reliability-Aware Representation Learning: integrating causal views into deep representation learning for better interpretability.
- Applied NLP under Noise and Imbalance: creating transformer-based systems for robust clinical text understanding.
What I'm Doing
Data Visualization
Designing intuitive and scalable visual tools that effectively turn complex datasets into actionable insights.
AI Community
Leading interdisciplinary collaboration to improve accessibility and practical AI adoption.
Cultural Leadership
Building impactful and inclusive cross-cultural programs and events with measurable, long-term impact.
Recent News
I am excited to announce that I have completed my Bachelor of Science in Applied Mathematics and Statistics, graduating with Highest Honors, at Emory University. I am deeply grateful to my mentor, Prof. Ruoxuan Xiong, for her guidance and support throughout my studies.
Our work “Learning Dynamic Representations and Policies from Multimodal Clinical Time-Series with Informative Missingness” has been accepted to the Findings of ACL 2026.
Our work “DART: Mitigating Harm Drift in Difference-Aware LLMs via Distill-Audit-Repair Training” has been accepted to ACL 2026 Findings.
Our work “MambaDATG: Domain-Adaptive Tri-Plane-Gated Pre-training for 3D Abdominal Segmentation” has been accepted as an oral presentation at ICASSP 2026.
My new personal website is now live! It is accessible worldwide, including regions where Google services are restricted.