Geoverdant

The hard part

Multi spectral scenes are far too large to process inside a request cycle. The processing service is separated from the API entirely, with background workers pulling NASA HLS tiles and computing vegetation indices band by band, so the frontend only ever asks for a finished heatmap.

The problem

Vegetation analysis over real land parcels means pulling multi spectral satellite scenes, aligning bands, and computing indices across rasters that are far too large to handle inside a normal web request.

The approach

I split the system: a dedicated FastAPI processing service with background workers owns imagery retrieval and raster math, while the product surface only ever requests finished, cached results.

What it does

• NASA HLS satellite imagery retrieval
• Multi spectral raster processing with Rasterio
• Vegetation index calculation across aligned bands
• Heatmap visualization generation
• Containerized processing pipeline

Architecture

A Next.js and NestJS product layer sits in front of a dedicated FastAPI processing service. Background workers handle NASA HLS retrieval and Rasterio based raster computation, results persist to PostgreSQL, and the whole pipeline runs in Docker.

Outcome

Turns raw multi spectral satellite scenes into vegetation heatmaps a non specialist can read, with the heavy raster work fully off the request path.