Case studies from current and recent engagements. Each names the problem, the constraint that made it hard, what I owned, and what was delivered.
Rebuilding malaria seasonality from one curve per country into an admin1-resolution, mechanism-aware emergence surface — water-balance and species-kernel models, calibrated and handed off to OpenMalaria.
read the case study
A Spectrum GOALS pipeline across 55 countries where the analysis and launch layer ran on the owner’s Mac and the heavy compute ran on a self-built Windows box over SSH — with Dropbox as a shared filesystem and results synced back.
read the case study
A variance-based global sensitivity analysis of Washington’s Environmental Health Disparities Map — measuring how much a census tract’s rank, and its “highly impacted” status, depends on defensible-but-arbitrary methodology choices.
read the case study
A 5 km geospatial model of solid-fuel cooking, the household air pollution it creates, and the under-five mortality attributed to it across 98 countries — a stacked ML ensemble feeding a Bayesian geostatistical model, published in Lancet Global Health.
read the case study