We train computer vision models on satellite and aerial imagery — and ship them to analysts, researchers, and operators who need answers from pixels, not slide decks.
Founded in Somerville, MA. AFRL co-development roots. CVPR-published research. Active B2G programs alongside a small set of commercial spinouts.
AFRL / SDL
Subcontract delivery · Penguin sites + BlackSky
NGA Accelerator
2023 inaugural cohort
Techstars Boston
2020–2021
CVPR
QFabric 2021 · GeoEngine 2022
21+ sources
Neso imagery pipeline
Universities
Paris-Saclay · Sorbonne · IIT Delhi · IIIT Delhi
Trusted by academia and government
What we do
DoD and intelligence work funded the platform. Disaster response, catalog discovery, and academic research keep it honest. A couple of commercial products reuse the same aerial vision models — different buyers, same underlying stack.
Industrial site monitoring, search-space reduction, and geospatial discovery prototypes delivered under AFRL and NGA programs. GeoEngine and Neso are the production stack underneath.
Operational basemaps and monitoring tools for wildfire, earthquake, flood, and spill events — built for teams that need current imagery interpreted fast.
GeoEngine, Neso, and GeoSearch — the geospatial ML stack we built for AFRL and NGA programs. Published at CVPR. The EarthAI developer portal is retired; the research and tooling continue in our gov work.
Two products share our aerial + vision AI stack but serve different buyers. Listed here for context — not what this site is selling.
Stack & spinouts
GeoEngine and GeoSearch are internal research infrastructure — not a public product. HADR and Inspect.Properties are the outward-facing tools.
Geospatial MLOps
End-to-end pipeline from imagery ingest to model deployment. Published at CVPR 2022. The core stack behind our AFRL and NGA deliverables.
Query geospatial assets in plain language
Natural-language interface over catalogs, orthoimagery, and model outputs — developed for our NGA gSEARCH proposal and gov programs.
Disaster monitoring and response
Basemaps, historical query, and alerts for wildfire, seismic, flood, and spill events. Built for humanitarian and defense-adjacent operational teams.
Property measurement from aerial imagery
Commercial spinout: satellite roof measurement, storm history, and claim documentation. Shares our vision models; separate product and team.
Research
QFabric
CVPR EarthVision Workshop 2021
Change detection on satellite imagery
GeoEngine
CVPR 2022
Geospatial machine learning operations
Timeline
Team
Siddharth Gupta
Co-founder · CEO
K L Bhanu Moorthy
Machine Learning Engineer
Rohit Kumar Singh
Operations
Supriya Dutta
Engineer
Chris Theall
Product
Dr. Sagar Verma
Co-founder · Advisor · fmr CTO
Dr. Matthieu Terris
PostDoc, INRIA
Dr. Maria Vakalopoulou
Assistant Professor, CentraleSupelec
Dr. Kavya Gupta
PostDoc, Saarland University
Research collaborations, gov programs, press, or just curious about the work — team@granular.ai
team@granular.ai