From Pixels to Predictions at Planetary Scale: RasterFlow Public Preview

RasterFlow let’s you solve complex monitoring challenges with vision-language models or tailored models for specific use cases, without needing to manage complex inference infrastructure.

It builds mosaics from multiple raster data sources, runs inference with computer vision models (onboarded or bring your own), and makes it easy to vectorize the results for downstream applications, through a high-level set of APIs.

Teams are already running RasterFlow at planetary scale:

• The USDA Forest Service uses RasterFlow to deliver wildfire containment predictions to its Wildfire Risk Management Division at an operational cadence.
• Miraterra uses RasterFlow to predict agricultural field boundaries across the U.S. Midwest and Canada’s Prairie Provinces, then joins detailed microbial samples and geospatial embeddings to those boundaries.
• Taylor Geospatial worked with Wherobots to produce 8.2 billion global field boundaries for the Fields of the World program.

What we’ll cover

• Predictable pricing. RasterFlow meters usage by the size of your input data, so you know the price of a task before running it. Pricing scales linearly, so you extrapolate from a small test to planetary scale.
• Your data, your storage. Tasks read from your own S3 buckets and write mosaics, predictions, and vectorized results back to them. Everything stays in open formats: Zarr, GeoParquet, and Iceberg through managed catalogs.
• Built-in visualization. Render mosaics, per-pixel prediction scores, and vectorized results on a single map right after a task finishes. With support for building and viewing Zarr stores optimized for visualization.
• Compared to Google Earth Engine. We’ll walk through the differences in pricing predictability, inference workflows, output access, and mosaicking cost for teams building on AWS.
• Using RasterFlow with Agents. LLMs and agents build RasterFlow notebooks and Wherobots Jobs through the Spatial AI Coding Assistant, and query results from Wherobots managed catalogs through our MCP server.