Intelligence for every acre.
YEZA is building geospatial AI that reads satellite, terrain, soil and weather data plot by plot — and turns it into clear guidance a farmer can act on.
Every field, measured from orbit.
Earth-observation satellites pass over Rwanda's hills every few days, recording light the eye cannot see. Healthy leaves reflect strongly in near-infrared; stressed or sparse crops do not. Repeated across a season, that signal becomes the raw material for every insight.
- OpticalSentinel-2 and Landsat multispectral imagery — crop development, land cover and change through the season.
- RadarSentinel-1 synthetic aperture radar — observations that continue through rainy-season cloud.
- TerrainElevation models — slope, aspect and drainage pathways across hillside farms.
Each farm becomes a place.
Pixels mean little until they belong to a real plot. YEZA anchors everything to a verified field boundary — with its crop, variety and planting date — so each satellite pixel, rainfall reading and soil sample is linked to a specific farm and a specific moment in its season.
Because the data model is field-centred, YEZA can compare zones within one field, fields within a cooperative, and patterns across districts.
Farmers and field agents can correct boundaries, crops and dates. The map learns from the people who know the land.
Pixels become indicators.
Raw reflectance is converted into layers an agronomist can reason with. Switch layers to see the same hillside four ways.
A change worth checking.
A single image is a snapshot. YEZA follows each plot through the season and compares it with its own trend and with nearby fields of the same crop and planting date. When part of a field falls behind, that zone is flagged — with the evidence and an honest confidence level.
View as table
Illustrative values. Gaps are cloud-masked observations — dropped, never guessed.
From a pixel to a farmer's phone.
A farmer doesn't need the map. They need a clear next step. Every YEZA alert is built to answer five questions:
- What was observed?
- Where in the field?
- Why does it matter now?
- What can the farmer do?
- How certain is the system?
YEZA: Maize plot 0412 — one area is behind its recent growth trend. Not a confirmed disease. Please check the marked area. Reply 1 for a call from your extension officer.
Ground truth makes the model local.
A model that works in one district, crop or season can fail in the next. Every field inspection, geotagged photo and measured harvest is linked back to its plot — with the farmer's consent — so YEZA is calibrated and validated in the places where it advises.
Each source is integrated only after its access terms, licence, resolution and update frequency are confirmed. National statistics give context; they never replace plot-level observation.
- Evidence beside every insight. Observation date, data source and confidence are always shown.
- Abstains when evidence is weak. No alert is better than a false one.
- Validated where it advises. Performance is reported by crop, district and season.
- Farmer data stays the farmer's. Informed consent, minimal collection, role-based access.
One crop. One district.
Real fields first.
YEZA is preparing a tightly scoped pilot: field mapping, satellite crop monitoring and alert workflows — checked on the ground before anything is scaled.