Granular.ai
From AltData.wiki, The Alternative Data Encyclopedia
Granular.ai is a United States geospatial artificial-intelligence company that develops computer-vision systems for satellite and aerial imagery. It works primarily on defence and intelligence, disaster response, and research infrastructure, while related commercial products apply aerial-image analysis to property and roofing workflows.[1]
The business operates as Granular Data, Inc. and identifies Somerville, Massachusetts, as its place of origin. Datarade independently lists Granular.ai as a United States provider of satellite and alternative data, although its automatically generated profile also assigns categories that are not explained by the company's current website.[1][2]
What It Does
Granular.ai trains and deploys models that identify patterns or objects in remote-sensing imagery. Stated government applications include industrial-site monitoring, reducing the area analysts must search, and geospatial catalogue discovery. Humanitarian and disaster applications cover wildfire, earthquake, flood, and spill events through operational maps, historical queries, and alerts.[1]
Products and Research Infrastructure
GeoEngine is described as an end-to-end geospatial machine-learning operations pipeline spanning imagery ingestion, model development, and deployment. Neso aggregates more than 21 imagery sources, and GeoSearch provides natural-language queries over image catalogues and model outputs. The company states that GeoEngine and GeoSearch are internal research infrastructure rather than public products.[1]
HADR AI is the outward-facing disaster-monitoring system, with named tools including FireMap, QuakeMap, FloodedLand, and SpillTrack. Inspect.Properties is a commercial spinout for roof measurement, storm history, and insurance or roofing documentation; roofing.io is presented as a separate contractor-operations business. These products share technical foundations but address different buyers.[1]
Data and Methodology
The company's technical model combines imagery acquired from multiple satellite and aerial sources with computer-vision training and deployment workflows. Granular.ai says Neso connects more than 21 sources and that government work has combined Penguin suborbital imagery with BlackSky monitoring. The available independent profile confirms satellite-data activity but does not verify the supplier count, model performance, or government deliverables.[1][2]
Granular.ai reports that QFabric, a change-detection dataset, was presented at the CVPR EarthVision Workshop in 2021 and that GeoEngine was published in a 2022 CVPR workshop. The specific publication URL tested during research was unavailable, so these publication details remain company-reported in this article.[1]
History and Programs
The company timeline says Granular Data, Inc. was incorporated in 2017, Granular.ai was founded in 2019, and geospatial AI co-development with the Air Force Research Laboratory began that year. It also records participation in Techstars Boston in 2020-2021 and the inaugural National Geospatial-Intelligence Agency Accelerator cohort in 2023.[1]
Those program milestones are described by Granular.ai; an NGA page attempted during research did not return accessible content. Datarade confirms the company's broad market presence but publishes no standard pricing and directs users to contact the provider. Public sources do not establish contract values, a complete customer list, universal model-accuracy figures, or current licensing terms.[1][2]