GeoInsight turns the entire Earth's surface into a unified vocabulary of Spatial Tokens, so AI can reason about geography the way language models reason about text.
Geography shapes everything: climate, conflict, supply chains, migration, disease. Yet most AI models treat location as a mere coordinate. We change that.
Every zone on Earth, from a city block to a continent, is represented as a Spatial Token in a hierarchical Discrete Global Grid System (DGGS). Tokens compose like words: local context stacks into regional meaning.
We encode multimodal Earth observation into one Spatial Token space — starting with Sentinel-2 from the 65 PB Copernicus archive. The analysis-ready token layer is live today; a foundation model trained on it is what we're building next.
Our goal: a model that can compare, forecast, and explain spatial relationships. Why is this zone at risk? What changed? Where should the next facility be built? Spatial intelligence as a foundation for decisions.
Any geospatial data source (rasters, vectors, point clouds, tabular) is mapped to, or natively stored in, the DGGS grid and converted into Spatial Tokens. The token hierarchy preserves multi-scale context automatically.
A standards-compliant OGC API DGGS endpoint exposes the token store for zone-based queries. Ask for a token, a region, or a spatial context window and receive structured, machine-readable answers.
The AI Earth Model will be trained on planetary-scale token sequences, learning spatial patterns, seasonal rhythms, and geographic relationships across decades of Earth observation data.
Bring your own labels and task definition. The model will adapt to agriculture, infrastructure, risk, logistics, or any spatially grounded problem — in days, not months.
GeoInsight is use-case agnostic — the same Spatial Tokens deliver answers across every domain. These are a few where spatial reasoning changes the game.
Dark vessel detection, chokepoint monitoring, and maritime pattern analysis. Spatial Tokens enable global ocean surveillance at scale.
Monitor land use change, deforestation, and ecosystem stress at global scale with per-token resolution.
Predict crop yield, detect stress early, and optimise field-level interventions across continents.
Site selection, expansion planning, and infrastructure risk assessment grounded in spatial context.
Rapid impact assessment and resource routing using real-time spatial reasoning over affected zones.
Physical climate risk, supply chain exposure, and geopolitical risk modelling with spatial precision.
Designed for cloud deployment from the ground up. Spatial Tokens are delivered as a scalable web service — no desktop GIS, no local downloads, no projection wrangling.
Built on the OGC API DGGS standard, interoperable with existing geospatial infrastructure from day one.
The same token hierarchy works at street level and continental scale, with no projection distortion.
Optical, SAR, thermal, LiDAR, and tabular data all live in the same token space.
Every answer traces back to specific Spatial Tokens and source observations. No black box.
Core DGGS infrastructure is open source. The model layer is where we build defensible value.
A team of engineers, geographers, and scientists who have spent careers transforming complex geospatial data into decisions — from extreme-environment expeditions to planetary-scale data infrastructure.


We're onboarding a small number of early partners. Tell us about your use case and we'll be in touch.