Home A · illustrative wind exposure
88 mph
Illustrative shelter factor: 88% of the incoming wind speed
Property-level wind risk analytics
Sharper wind risk at the building level.
Downstream delivers hyperlocal wind intelligence on top of the catastrophe models the market already runs, helping underwriters see where building-level risk diverges from modelled expectation.
Property wind lab
100 mph wind
Same buildings. Same incoming wind. Different surroundings.
Wind west to east
Illustrative comparison, not a wind simulation or loss prediction. Both buildings have the same construction. Tree shelter varies with wind direction, tree height, spacing, and condition; the reduction shown is an example, not a measured estimate.
Home B · illustrative wind exposure
100 mph
No tree shelter: 100% of the incoming wind speed
The problem
Cat models price hurricanes at a kilometre. Losses happen at a building.
Two properties one street apart can carry very different true wind risk. Roof form, elevation, tree cover, nearby structures, and channelled wind flow all matter, but those signals rarely fit cleanly into legacy exposure schemas.
~1 km
Typical native hazard resolution
850k+
Labelled buildings in public damage datasets
The solution
One multiplier. Zero change to your stack.
Downstream learns where observed damage diverges from modelled expectation, then delivers that difference per property.
Your model output
Your existing model supplies the baseline loss estimate.
Property-level wind-damage foundation model
We start with the property.
Identify the building and its characteristics in property-level imagery.
NOAA post-storm imagery · Hurricane Ian
Then we look around it.
Nearby buildings shape a property’s immediate surroundings and wind exposure.
NOAA post-storm imagery · Hurricane Ian
Then we bring in the wind.
Connect the property to the storm’s wind field and event history.
Hurricane Ian · modelled gusts (left), radar reflectivity (right)
Learn from what happened.
We combine property, context, and wind information using machine learning. Our training targets are proprietary damage labels covering historical hurricanes across the past decade.
Property-adjusted view
Apply a property-specific multiplier to your baseline estimate to support selection, pricing, and portfolio decisions.
Founding team background
Insurance, catastrophe science, and ML under one roof.
The founding team brings training and operating experience from leading research, aerospace, and risk analytics institutions.
Get in touch