Trained and evaluated a semantic segmentation model (5.78M parameters) for satellite land-cover classification, and built a complete Streamlit application around it:
On-demand Sentinel-2 imagery retrieval for protected area shapefiles
Mask-based change detection on an interactive Streamlit map
Geospatial Deep Learning: Land-Cover Classification
PythonCNNrasteriogeopandas
Built and hyperparameter-tuned a CNN for satellite land-cover classification, and a geospatial pre-processing pipeline to extend it to a real, unlabelled region (Kajiado County, Kenya).
Built an interactive Tableau dashboard exploring electricity pricing, affordability, and the renewable-energy transition, synthesising government tariff schedules, household income data, and regional generation statistics.
Full-time
UX Researcher
Jireh Health
Apr ’25 – Sep ’26
Accessible healthcare-financing web app, serving 4 facilities and 1,800+ users.
User ResearchPrototypingAccessibilityMobile UI
Sole designer working from PM-defined acceptance criteria through to engineering handoff.
Redesigned flows to:
cut median signup time from 5.2 to 2.1 mins (-59%).
Led Jireh Health’s frontend design-system modernisation by:
introducing reusable primitives, semantic tokens, and consistent UI patterns
improving scalability and maintainability across mobile and web experiences.