PhD from Texas A&M University
I develop algorithms that help machine learning models adapt to new settings with stronger performance and lower computational costs. My work spans parameter-efficient fine-tuning, physics-informed modeling, and model evaluation, including settings with limited labeled data and noisy signals. I apply these approaches to digital health, computer vision, acoustics, and sensing.
Publications
IEEE International Conference on Bioinformatics and Biomedicine · 2026 · ACCEPTED
Frozen pathology encoders and small adaptation modules predict ethanol drinking status from nonhuman primate liver slides, reaching 78.1% mean balanced accuracy with four training subjects per class.
IEEE International Geoscience and Remote Sensing Symposium · 2026
Learnable histogram modules adapt frozen Transformers to passive and active sonar classification, reaching 91.8% mean accuracy on VTUAD with 153,000 trainable parameters.
IEEE Sensors Journal · 2022
A low-power ECG/EDA sensor combines physiological measurements with machine learning, achieving 94.4% mean stress-classification accuracy in an 18-participant study and up to 70 hours of battery life.