Spatial Data in Apache Iceberg: Optimizations and Management That Matter

Spatial data in Apache Iceberg needs different optimization than tabular data. A geometry column has no natural sort order, so unsorted files carry wide, overlapping bounding boxes and query planners cannot prune them… At all… This behaviour turns a selective spatial filter into a full table scan. A second problem compounds it: one oversized geometry […]

Spatial Graph RAG for the Physical World

Introduction RAG (Retrieval Augmented Generation) has addressed one of AI’s biggest challenges for enterprise users: missing or hallucinating empirical business and real world context . Instead of generating answers from nothing, RAG retrieves relevant documents and feeds them to the model as context. It works. Ask an AI about your company’s Q4 revenue, and RAG […]

Building the Wherobots Mobility Solution Accelerator: A Technical Deep Dive

Three Notebooks, One Medallion Architecture, Full 4D GPS Trajectory Processing: Part 2 of 2

How well does SAM3 detect building footprints? We asked the Wherobots Spatial AI Coding Assistant

In a recent post, we showed how easy it is to use RasterFlow and Meta’s Segment Anything 3 Model (SAM3) to detect features in the physical world. A single end-to-end pipeline built a 133 GB NAIP mosaic of Marion County, Oregon, ran SAM3 against it with text prompts spanning eight classes, and produced approximately one […]

Wherobots MCP Server: Building GEOINT Spatial Pipelines with AI Agents

I built three national-security GEOINT use cases on the Wherobots stack in days instead of weeks. A Critical Infrastructure Vulnerability (CIV) pipeline with two regional variants, plus a border-corridor analysis on real transportation segments. The Wherobots geospatial MCP server is what made that timeline possible. Most of the work in standing up a credible use […]

Change Detection Using AlphaEarth Foundations (Part 2)

Change detection with AlphaEarth Foundations measures how a place changes over time directly in embedding space, without hand-built spectral indices. This post scores temporal change using leave-one-out medoid distance and robust z-scoring per pixel, then runs it at scale on RasterFlow with a Colorado demonstration. The output flags where the landscape shifted and how strongly, ready to vectorize into geometries in WherobotsDB.

AlphaEarth Embeddings, Zonal Statistics, and PCA

Aggregate AlphaEarth embeddings over Iowa fields and visualize them with PCA.