AI for Good Lab

Weather Forecasting

We develop next-generation AI-based weather forecasting models that complement and enhance traditional numerical weather prediction (NWP) systems. Our research focuses on learning atmospheric dynamics directly from large-scale reanalysis and observation datasets, enabling fast, accurate, and scalable global forecasts.

Kolmogorov Arnold Network

We study Kolmogorov–Arnold Networks (KANs), a function-based neural architecture that replaces fixed activations with learnable univariate functions, enabling interpretable and expressive modeling of complex physical systems.

Generalizable Deep Learning for Scientific Domains

Our research methodologies are broadly applicable and include remote sensing analysis, medical dataset classification and regression, and super-resolution, enabling cross-domain transfer of advanced AI techniques.