Science

Model outputs inherit every choice made upstream.

I work on the assumptions, inputs, and computational systems behind disease models: HIV and malaria portfolios, allocation across countries and populations, geospatial exposure, and composite indicators.

Here, national curves become subnational profiles, serial model runs become parallel batches, and public-health rankings are rebuilt under alternative methods.

Currently

A shared assumptions layer for the IPM malaria portfolio (BMGF)

Selected work

Malaria PST Replacing national malaria curves with 362 subnational profiles Geospatial methodology Water-balance and mosquito-species models tuned against MODIS NDVI phase and exported as OpenMalaria inputs. Read the case study IPM HIV Parallelizing 500+ HIV modeling branches across 55 countries Systems + compute One-at-a-time Spectrum runs became full-portfolio, scenario, and sensitivity batches on a multicore Windows box. Read the case study WA EHD Map Stress-testing Washington’s Environmental Health Disparities Map Sensitivity analysis 5,000 index rebuilds measured which methodology choices moved tract ranks and rank 9–10 classifications. Read the case study IHME HAP Estimating solid-fuel cooking at 5 km across 98 countries Geospatial modeling Published estimates of household-air-pollution exposure, attributable under-five respiratory mortality, and progress toward the 2030 target. Read the case study

All work

Deliverables

Docs & reports from client engagements

Elsewhere

CV · GitHub · Google Scholar · LinkedIn

Contact

Open to select engagements

Model review, sensitivity analysis, and the data systems under them.