I build the data side of impact measurement: monitoring and evaluation (M&E) systems, MEAL workflows and donor-reporting automation for United Nations and NGO programmes in East Africa. The work ranges from humanitarian analytics platforms that join UN open data, to production ERP/POS systems that run Kenyan retail, to applied AI — RAG systems and profiling engines — grounded in authoritative frameworks. What holds the portfolio together is a governance-first approach: auditable pipelines, partitioned data sovereignty, reproducible analyses and documented trade-offs, so that every claim a system makes can be traced to evidence.

gcf-food-security-intelligence — the flagship data-science platform: production-grade, end-to-end prediction of food-insecurity risk with climate-scenario simulation. It packages ingestion, feature engineering, model training and early-warning outputs into one governed Python codebase, built so that a food-security analyst can move from raw indicators to a ranked risk view without leaving the repository. The full formal title — Global Climate & Food Security Intelligence Platform — lives in the README and description; the repository carries the compact working name.

mbumah-hardware-pos — the flagship business system: a multi-tenant hardware-store point-of-sale and ERP in TypeScript. It models the full retail loop of Kenyan hardware retail — catalog, suppliers, sales, payments, stock and reporting — per tenant, and is the most operationally mature platform in the portfolio, carrying the account's richest topic coverage and active deployment.

un-review-hub — a federated human-in-the-loop review platform designed for UN agencies. Its defining engineering decision is data sovereignty through partitioning: each agency's review data stays isolated while review workflows remain unified, which is the property that makes cross-agency review lawful in practice. It is the clearest statement of the account's governance-first architecture style.

sdg-rag-system — SDG-Trans-RAG, a retrieval-augmented generation system that grounds urban-transport intelligence in UN-Habitat SDG 11 frameworks and multi-modal transport data. It demonstrates the account's applied-RAG craft: retrieval over authoritative framework documents, fusion with operational data sources, and answers that stay citable and auditable.

kenya-malnutrition-risk-prediction — an open-source digital tool predicting acute childhood malnutrition risk in Kenya with machine learning. It is public-health data science in the account's home market: survey and climate indicators in, ranked risk out, with the full formal system name preserved in the README. The repository was smart-shortened from a 52-character title-case chain to a readable kebab-case name.

Account infrastructure — this profile index and the private Zero-Red Mandate governance ledger.

Legacy archive — cv kept public as the 2022-era CV record; the learning-era repositories (android, android-ui, jupyter-predictions, trail-one, newsletter-popup-project) are archived with context stubs.

Available for M&E data-systems work: platform builds, evaluation-support automation, donor-reporting pipelines and humanitarian data engineering across East Africa. The fastest routes in are the portfolio site and X / Twitter — or open an issue on any public repository above and it will find me.