Research Projects

In close collaboration with our industry partner RaSpect, we are developing AI- and foundation-model-driven approaches to assess infrastructure defects in drone imagery. These methods improve transferability across diverse settings and robustness to noisy annotations, enabling foundation vision models such as the Segment Anything Model to specialize in crack segmentation with minimal annotation cost.

Tropical cyclones (TCs) cause significant financial losses and threaten millions of people living along the world's coastlines. Growing coastal populations and climate change are increasing this vulnerability. We develop neural-parameterized, GPU-accelerated methods that rapidly and accurately quantify uncertainty in tropical-cyclone and storm-surge risk.

We explore how UQ methods can enhance AI models by rigorously assessing the confidence in their predictions and improving their robustness, particularly when dealing with noisy, incomplete, or novel data. Concurrently, we leverage powerful AI techniques, such as deep learning and surrogate modeling, to tackle the computational challenges often associated with traditional UQ methods, making complex uncertainty analysis more feasible. Ultimately, this project aims to develop next-generation AI systems that are not only accurate but also reliable and trustworthy, capable of quantifying their own uncertainties for safer and more informed decision-making in complex real-world applications.

This research develops and applies physics-informed machine learning (PiML) to improve the modeling of wave and wind energy systems. By embedding physical laws in model architectures and training, we aim to combine the reliability of numerical simulation with the efficiency of data-driven methods. The resulting surrogate models predict wind and wave interactions with turbines and energy converters, accelerating the design, optimization, and control of renewable energy technologies.