GroundShift

Built the full detection pipeline for groundshift today
satellite imagery goes in, anomaly alerts come out.

what shipped:
→ preprocessing pipeline cloud masking, co-registration,
Z-score normalisation
→ NDVI-based change detector computes per-pixel change
scores between two satellite scenes
→ alerting layer; converts change scores to GeoJSON polygons,
fires webhooks when anomalies exceed threshold
→ 24 tests, all passing

the interesting bit: groundshift doesn’t just flag change
it scores how anomalous that change is against the historical
seasonal baseline for that exact location. a field turning
brown in october is normal. the same field in april is an alert.

next up: building the UI–> a full geospatial intelligence
dashboard. dark map, live anomaly overlays, alert feed.
think mission control, not a settings page.

Started building groundshift — an open source Python framework
for satellite change detection. The idea: point it at any polygon
on Earth, get alerted when something changes. Deforestation,
floods, conflict damage — all from free ESA Sentinel satellite data.

Day 1: built the full ingestion and preprocessing pipeline from
scratch. Sentinel-2 scenes now download automatically from the
Copernicus archive, get cloud masked, co-registered, and
normalised. 13 tests passing. Detection layer (NDVI-based change
scoring) is next.