Wherobots

🔥🔥🔥 On September 21, 2026, California declared an El Niño emergency, and the debris-flow hazard map stops at the bottom of the hill. The houses start there. Above Altadena, 7,718 mapped buildings sit inside or within 500 m (1,640 ft) of the 42 Eaton burn-scar basins that USGS rates high hazard. WherobotsDB counts them in one SQL engine, joining the USGS basins, a 30 m elevation grid, and Overture buildings and roads where they sit. The elevation join adds 411 km (255 mi) of road on the low ground below the scar. Read it: https://lnkd.in/gpez7bBt Try Wherobots: https://lnkd.in/gAP5VhMZ #SpatialSQL #GeospatialAnalytics #ElNino #Raster #Vector

Looking forward to the CNG forum in Snowbird! Join our Workshop with Matt Forrest (my guy from Wherobots) and Chris Stoner from Amazon Web Services (AWS). We build with Kiro + Wherobots Spatial AI coding tools. Output: a risk intelligence layer you can tune by industry. We make the hard bits easy! https://lnkd.in/g9hH55hu

Converting satellite imagery, vector data, and portfolio holdings into location-level risk scoring, at low latency and at scale, is the hard part of geospatial risk work. Amazon Web Services (AWS), Felt, and Wherobots are co-hosting a hands-on virtual workshop to build that pipeline in one session. Architect and deploy a geospatial agentic AI stack and leave with the blueprint: • Spatial joins and risk scoring across satellite imagery, building and parcel geometry, and portfolio holdings, on Amazon Aurora PostgreSQL with PostGIS and Wherobots • AI agents orchestrating it end to end with Amazon Strands, Amazon Bedrock, and Kiro IDE • A transparent, weighted risk-scoring methodology, visualized on an interactive Felt map The use case is risk intelligence: triaging insurance exposure after a catastrophe, assessing risk to critical infrastructure, quantifying weather-driven disruption for an investment thesis. Teams run this across financial services, insurance, energy, agriculture, and government. Built for data engineers, solutions architects, GIS analysts, and developers working with location data. Come ready to build live. Presenters: Rajeshkumar Sabankar, Sarabjeet Singh, Damion Harrylal, Ben Pruden, Jaime Sanchez, Pranav Toggi Thanks to Eliza Connors for pulling this together. Thursday, October 8 at 9AM PT | Virtual Register: https://lnkd.in/gWgyvz5x

3 big things happened in earth observation last week. Each one hits a different part of the EO toolkit. Sensors: ESA's FLEX and Copernicus Sentinel-3C went up together. FLEX measures the fluorescence plants give off during photosynthesis helping get more direct data on plant health. It flew in tandem with Sentinel-3C. Inference: Wherobots opened RasterFlow to public preview. Mosaics, models like SAM3 and Fields of the World, and vector output to GeoParquet in your own cloud bucket. You can price a job before it runs: for example very solar array in a 500 km² county from a text prompt ran for about $35. Search: LGND AI, Inc. lets you query satellite imagery in plain language. Type "construction site with heavy machinery" and get matching locations back, built on Clay embeddings. No labeled training set. For years the hard part of EO was getting clean, analysis-ready pixels. Now the pixels arrive, inference is a function call, and search is a sentence. The hard part is knowing which question to ask and what to join the answer to. That's a spatial thinking problem, but the toolkit is catching up fast. 🌎 I'm Matt Forrest and I talk about modern GIS, earth observation, AI, and how geospatial is changing. 📬 Want more like this? Join 15k+ others learning from my daily newsletter → Link in my profile

When the execution engine is loosely coupled with orchestration, the pipeline is blind while waiting for one fact: whether the job finished, failed, or died. That was the case for Wherobots Cloud jobs and the AWS pipelines launching them. Daniel Smith designed a reference implementation using AWS Step Functions, the callback pattern, and the Wherobots Python SDK: four Lambda functions, two state machines, one API Gateway endpoint, and a Wherobots job script. Every AWS resource is created by plain CLI calls you can read. A guided app drives the whole lifecycle: upload, deploy, run, teardown. The demo counts Overture Maps buildings within one kilometer of downtown Seattle (geodesic buffer, EPSG:4326) and returns 1,084. Full post: https://bit.ly/4yOL5fI

Physical world AI keeps running into the same wall: the world is not in the data. We measured one gated townhome community with five queries against open map data. - 318 delivery addresses, 14 mapped building footprints - 18 separate addresses carry the same unit number, the nearest 40 metres apart - No mapped walkable path comes within 33 metres of the door, though a road reaches 20 The unit records exist. The geocodes are correct. Deliveries still arrive at the wrong door. A human absorbs this. They read placards, walk the row twice, call the customer. That improvisation is unpriced error correction, and it is the only reason these numbers have not already broken last-mile delivery. Remove the human and the absorption goes to zero. A sidewalk robot needs a traversable path. A drone needs a classified landing surface. Each of them needs the destination as geometry. "Unit 108" is a string. Physical world AI is a context problem before it is a model problem. Read it: https://lnkd.in/eeEwqvN8 #PhysicalAI #LastMile #AutonomousDelivery

🚢🚢🚢 Traffic through the Strait of Hormuz shipping lanes fell 95% after the February closure. Vessel counts at Iran's Kharg Island terminal and the Fujairah anchorage outside the strait held at or above 2025 levels. Many ships around the strait switched off or falsified their transponders (AIS). Steel hulls reflect near-infrared light either way, and WherobotsDB counted hulls across 719 free Sentinel-2 scenes, reading about 1.5 GB of their 151 GB. - A near-infrared threshold found the hulls with no trained model - Overture Maps land polygons removed islands and coastline - The whole analysis cost about $63 Read it: https://lnkd.in/g8ErtFSS Try Wherobots: https://lnkd.in/gAP5VhMZ #EarthObservation #Raster #Sentinel2 #RemoteSensing #GeospatialAnalytics #Wherobots

The USDA Forest Service is using RasterFlow to deliver wildfire containment predictions at an operational cadence. Taylor Geospatial worked with Wherobots to produce 8.2 billion field boundaries on it. It's now in public preview. Miraterra is also running it across the U.S. Midwest and Canada's Prairie Provinces for agricultural field boundaries. RasterFlow runs vision models on planetary-scale raster data without the inference infrastructure overhead. Build mosaics, run inference with onboarded or bring-your-own models, and vectorize results through one set of APIs. Join Philip Darringer and Ryan Avery live for a walkthrough. What's on the agenda: • Pricing metered by input data size, with linear scaling from a small test to planetary scale • Reading inputs from and writing outputs to your own S3 buckets in Zarr, GeoParquet, and Iceberg • A side-by-side with Google Earth Engine for teams on AWS • Building notebooks and Jobs with agents through the Spatial AI Coding Assistant and MCP server Wednesday, September 30 at 10 AM PDT 👉 Save your spot: https://bit.ly/4xjlPNz

Our latest announcement at Wherobots: RasterFlow is now in Public Preview! RasterFlow is a serverless engine for planetary-scale Earth observation with a high-level Python API for building mosaics, running inference with computer vision models, and vectorizing the results into GeoParquet geometries. With a predictable pricing model based on the size of the data processed, you can accurately estimate the cost of any task before execution. Teams are already using RasterFlow at planetary scale. Taylor Geospatial produced 8.2 billion global field boundaries. The USDA Forest Service runs its wildfire containment model across the western US at a daily cadence. And Miraterra uses RasterFlow to predict agricultural field boundaries, then join detailed microbial samples and geospatial embeddings to those boundaries. Read more: https://lnkd.in/gWd9PF7u Sign up today to try it out: https://lnkd.in/gNF-_wDi