This research project demonstrates how modern data science, GIS, Python programming, machine learning can process, fuse, and analyze heterogeneous claims and meteolorlogical data at multiple scales. Applying core techniques from spatial data engineering, and applied statistics, students will build reproducible workflows to clean noisy claim records, geocode and validate building locations, and integrate building-level losses with 1 km rainfall and wind fields. They will explore patterns through machine learning, mapping and visualization, design features that capture thresholds and extremes, and develop predictive models that relate meteorological drivers to building loss. Through rigorous cross-validation and uncertainty analysis, students will assess model skill and communicate results with clear figures, concise memos, and well-documented code.