Food security · Survey harmonisation

Config-driven, tested workflow harmonising 14 rounds of Chad's national food security survey (ENSA, Oct 2018 to Oct 2025; 197,947 households) into comparable thematic datasets: food consumption, coping strategies, expenditure and more. Indicators are recomputed and validated against the original analysts' scores, with provenance manifests.

Earth observation · Species distribution modelling

R package with a Shiny interface that makes species distribution modelling accessible to non-programmers: raster predictors and occurrence data in, spatially blocked cross-validation (blockCV), predictor screening (CENFA), MaxEnt and other models out, plus the generated R code for full reproducibility. Example data: agroforestry tree species in Niakhar, Senegal.

Public health surveillance · Impact evaluation

R pipeline on the weekly IDSR surveillance database for Ituri province (DRC): automated weekly epidemiological bulletin in Word, a French-language dashboard (diseases, health zones, malaria, Ebola, data quality) and a controlled interrupted time series analysis of routine morbidity and Ebola burden across 36 health zones. Runs end to end on synthetic data with Rscript run_all.R --demo.

Public health surveillance · Forecasting

Ten years (2015 to 2024) of Niger's weekly notifiable disease reports cleaned into tidy, source-validated tables and served in an R Shiny dashboard: disease trends, meningitis alert thresholds, reporting completeness, maternal deaths and a DHIS2 comparison. A modelling module fits negative binomial, hurdle and endemic-epidemic models with forecasts.

Earth observation · Agricultural land

Google Earth Engine workflow that maps arable land at 30 m resolution across African countries by combining cropland extent (NASA GFSAD 2015) and cleared forest, then masking steep slopes (>15%), permanent water and impervious surfaces. One script per country, from Burkina Faso to Uganda.

Teaching · Spatial data science · Course website

Eleven practical sessions taught at ENSAE Dakar: vector and raster data, zonal statistics, risk mapping, combining health, WorldPop and ACLED conflict data into composite indicators, spatial autocorrelation, species distribution modelling and Sentinel-2 spectral indices (NDVI, NDWI, NDBI). In French, CC BY 4.0.

AI agent · Food security surveys, Chad

Ask how the 14 ENSA rounds were harmonised and what they show. 15 read-only tools trace any variable across rounds, explain recodes and quality checks, compare 58 indicators by round, region and sex of the household head, and draft crosswalk mappings for a new round. Regional cells under 30 households are suppressed.

AI agent · Resilience programme surveys, Chad

Ask how the 10 survey rounds of WFP's Integrated Resilience Programme (17,255 households) were merged, which validation checks fail, which data issues remain open, and what 55 indicators show by round and region, with WFP's reconciled costs. The server refuses charts that join lean-season and post-harvest rounds. Evaluated on 18 questions with expected values and caveats.

AI agent · Disease surveillance, Niger

Ask what Niger's weekly notifiable disease surveillance shows: trends, seasonality, like-for-like year comparisons, meningitis epidemic thresholds, reporting completeness and model estimates with confidence intervals. 19 read-only tools enforce the rules an epidemiologist would: partial years are never compared with full years, missing reports are never read as zero, and case-definition changes are always flagged.

AI agent · Epidemiological surveillance, DRC · work in progress

Question-and-answer agent on Ituri's IDSR surveillance data, packaged as a Claude Code skill. Indicators are computed by the R pipeline, never by the model, and a parity test checks the tools against the pipeline across diseases, zones and weeks. Evaluated on 30 questions against answers computed independently in R. Real data stay locked until a written data-sharing agreement is in place.

Impact evaluation · Cost-benefit analysis · 2018 to 2023 · Report

Household dataset for IFPRI's cost-benefit analysis of WFP's Integrated Resilience Programme in Chad: 10 survey rounds, 17,255 households, 626 columns from 26 modules, WFP costs reconciled by activity, a Shiny dashboard of checks and results, and a package for public release on IFPRI Dataverse. Known defects are documented, and each fix writes a new versioned file.

Outbreak epidemiology · Health surveillance · 2026

Reproducible pipeline for the MSF Ebola treatment centre in Mongbwalu, Ituri: weekly epidemiological sitrep, line-list analysis report, a daily data quality checklist for the data entry team, and DHIS2 analyses (epidemic curves, transmission chains, Rt). Patient data never leave the analyst's machine; CI and demos run on synthetic data. Code © MSF OCG; reuse requires agreement.

Conflict · Climate · Spatial exposure measures

Exposure measures for survey clusters (10, 25 and 50 km buffers) and the 519 Admin 3 units of the DRC: ACLED conflict events 1997 to 2025, WorldPop population and age-sex structure 2015 to 2025, and ERA5-Land temperature anomalies 1991 to 2025, with an ACLED versus UCDP GED validation. Outputs in CSV and Stata.

Food systems · Labour · Data acquisition · in progress

Data acquisition for the Mastercard Foundation agrifood scenarios study with IFPRI: a verified source matrix and access register for the labour force and household surveys of 17 countries, with scripts that validate the register, build the Excel deliverables, download from the World Bank Microdata Library and block any microdata from being committed.

Shocks · Coping strategies · 2019 to 2023

How households respond to conflict, climate and price shocks, using eleven waves of cross-sectional household survey data linked to an extensive set of shock indicators compiled from external sources.

Shocks · Coping strategies · 2018 to 2023

Companion analysis for Mali using association rule mining to identify which coping strategies households combine when facing different kinds of shocks and stressors.

Conflict · Spatial indicators

Implementation in Google Earth Engine of ACLED's conflict diffusion indicator: share of populated 10 km grid cells with high levels of violence, built from WorldPop and ACLED and aggregated to admin 2, with sensitivity analysis on the population and event thresholds. Prepared for WFP.

Poverty mapping · Small area estimation

Poverty headcount (FGT0) below the levels where the EHCVM 2021 survey is representative (provinces and communes), comparing direct, Fay-Herriot and machine learning estimates with conflict and climate covariates. Part of a series also covering Senegal, Mali, Niger, Benin and Côte d'Ivoire.

Food systems · Data engineering

Configuration-driven, provenance-tracked pipeline assembling 30 sources across five layers (geography, soil, food, health, policy) into one database keyed on a county and sub-county crosswalk, built for soil-to-nutrition pathway analysis. Every object carries publisher, licence, checksum and extraction date.

Earth observation · Agriculture · replication guide

Guide to replicating the Fields of The World global field boundary map at 10 m for Burkina Faso: cloud-free Sentinel-2 mosaics, inference with the PRUE U-Net model, polygon extraction and confidence filtering, for seasonal updates or local fine-tuning.

Offline-first field data collection · Phase 1

Platform for collecting field data in low-connectivity areas, built by extending ODK: ODK Central hosted on AWS in Africa (af-south-1), a branded ODK Collect, and an encrypted SMS fallback when there is no data connection. The SMS codec and the bridge service into ODK Central are built and tested; the infrastructure is written and validated but not yet deployed.

Survey supervision · Senegal · in French

Supervision tools for the ANSD survey of beehive productivity in Senegal, collected on tablets with Survey Solutions: fieldwork progress by day, region and enumerator, 39 consistency checks, and a ready-to-use list of interviews to reject. An R Shiny app and a Python script give identical results.