Nationwide composite risk model across all 3,144 US counties, combining grid reliability metrics, weather hazard exposure, elderly population share, and poverty rate into a single weighted vulnerability index.
Spatial regression analysis testing whether police response times in Detroit correlate with neighborhood race and income, mapping the geography of delayed emergency response across 139 census tracts.
XGBoost classification model predicting which of Chicago’s ~800,000 unverified properties are likely connected by a lead service line, using Cook County Assessor records to prioritize infrastructure replacement.
Air quality exposure analysis across 25 LA school bus routes measuring PM2.5 and NO2 at 100m intervals, revealing students in low-income neighborhoods breathe 2.4x more polluted air on their way to school.