Granular.ai

Market position

Granular.ai is an early-stage, small-team geospatial AI vendor whose credibility markers are government program selections (AFRL and SDL grants, the inaugural 2023 NGA Accelerator cohort with a stated $100,000 grant, Google Cloud TPU Research Program support), Techstars Boston participation, CVPR-published research and academic partnerships. Aggregator listings name Atlas AI, DataRobot, Gravity AI, Destined AI, SpaceKnow, Groundlight AI, Mach9 and Landing AI among similar companies. Buyer-facing coverage flags its scale and stability as the main open question for programs with long timelines.

The company positions GeoEngine as an end-to-end geospatial machine learning pipeline spanning ingestion, annotation, training and deployment in a single platform, a scope that a third-party review describes as covered end-to-end by few geospatial AI products. Additional differentiators cited include a 21-source imagery pipeline, natural-language catalog querying via GeoSearch and Geo LVM, peer-reviewed research published at CVPR (QFabric 2021, GeoEngine 2022), and a record of US defense and intelligence program selection (AFRL, SDL, NGA Accelerator) plus an open-source urban change model on a domain-verified GitHub organization.

Technology

The stack centers on computer vision models trained on satellite, aerial and drone imagery. GeoEngine handles the full model lifecycle — data sourcing, AI-assisted annotation, training, deployment as APIs, and performance tracking — and ingests optical, SAR and drone imagery. Neso aggregates 21 or more satellite imagery sources, with Sentinel-2 named publicly and used in FireMap. GeoSearch layers a natural-language query interface over catalogs, orthoimagery and model outputs, powered by Geo LVM, a multi-modal large language and vision model. Supporting components described in third-party coverage include DL Judge (model compression), Synthetix (synthetic geospatial training data) and Europa (dataset development). Published research covers change detection (QFabric, CVPR EarthVision Workshop 2021) and geospatial MLOps (GeoEngine, CVPR 2022), with an open-source "fabric" urban change model released on GitHub.

Go-to-market

Sales are direct and quote-based, with no published pricing; inquiries are handled through a contact form and email. The government track runs through program participation and subcontracting — AFRL/SDL subcontract delivery, NGA Source Foundation ecosystem work and the NGA Accelerator — while research reach comes through CVPR publications, academic collaborations and an open-source model release. The commercial spinouts Inspect.Properties and roofing.io are run as separate products with their own go-to-market, with Inspect positioned as a self-serve analytics tool for insurance, construction and real-estate buyers.

Primary buyers are US defense and intelligence organizations (AFRL, SDL, NGA) and government program teams, plus humanitarian and defense-adjacent disaster response operators. Research users include academic groups at Paris-Saclay, Sorbonne, IIT Delhi, IIIT Delhi and IIT Gandhinagar. On the commercial side, Inspect.Properties targets insurance carriers, adjusters, construction firms and real-estate users, and roofing.io targets roofing contractors. A review characterizes the two buyer types as R&D teams building custom geospatial vision models and operational end-users consuming finished analytics.

Geography

The company describes itself as founded in Somerville, Massachusetts, and a third-party review lists an address at 240 Elm Street, Somerville, MA. A Techstars job-board listing associates the company with multiple US locations including Somerville, MA; Belmont, VA; Somerville, NJ; Belmont, MA; and Massachusetts more broadly. Research collaborations extend to universities in France and India, including Université Paris-Saclay, Sorbonne University, IIT Delhi, IIIT Delhi and IIT Gandhinagar. Its imagery and analytics coverage is global in scope.

History

A predecessor entity, Granular Data, Inc., was incorporated in 2017; the company dates the founding of Granular.ai to 2019, when co-development of geospatial AI with AFRL began. It joined Techstars Boston in 2020, the same year it built the Neso imagery pipeline and ran a Helios Towers proof of concept. Research milestones followed with QFabric at the CVPR EarthVision Workshop in 2021 (alongside Penguin site and BlackSky monitoring work for AFRL) and GeoEngine at CVPR 2022, the year it pivoted away from steel PMI discovery. In 2023 it was selected for the inaugural NGA Accelerator cohort, submitted the gSEARCH proposal, and introduced the Inspect.Properties brand. Its 2024-26 timeline entries list active government programs, Inspect go-to-market, the roofing.io spinout, and retirement of the EarthAI developer portal. Third-party listings vary on founding year (2017 per aVenture and the Techstars job board) and on headquarters and status.

Risks & controversies

Third-party review coverage notes that pricing is not published, requiring a direct sales conversation for budget clarity; that no named customer count or confirmed customer numbers are disclosed; and that the company neither operates satellites nor publishes a formal data licensing agreement, so buyers are advised to request data provenance and sub-licensing terms for each of the 21 ingested imagery sources. The same review raises scale and stability as considerations for a team of roughly five people, and notes no total funding figure is published and no formal institutional VC round had been completed as of mid-2026. Source records also conflict: one aggregator lists founding in 2017 and an operating status of "acquired" with no acquirer named, while the company's own timeline dates its founding to 2019 with the 2017 date applying to predecessor entity Granular Data, Inc.