Rafy's research focuses on understanding how hydrologic models represent the physical processes that generate streamflow and flooding and inferring their controlling factors, using a combination of physics-based reasoning and machine learning to build more trustworthy models for water resource and hazard prediction.