LandFall
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Landfall is an AI agent that does the first damage count after a disaster. Before a county can get federal aid, it has to show FEMA how many homes were damaged and how badly. Today that means teams driving street by street for days, while families wait. With Landfall, you draw a box over a town and in about a minute the agent does that first pass by itself. It finds the before and after aerial photos (NOAA and USDA in the US, Maxar abroad), lines up every building, measures lost roof and structure with SAM 3.1, estimates flood depth at each house from surveyed high-water marks, and drafts a FEMA damage category for every building with Muse Spark Open. Then it checks its own work: drafts that don't fit the measurements are questioned, and hard calls go to Muse Spark for a second opinion. It ranks which buildings to visit first, estimates repair costs, and makes a printable claim page for every building in English, Spanish or Haitian Creole. Anyone can ask questions by typing or by voice. It works for hurricanes, wildfires, floods and earthquakes, and people can upload their own aerial photos. On 50 Cedar Key buildings after Hurricane Helene, the open model alone matched our by-eye labels on 47 and 48 of 50. Every result is marked as a draft for a trained assessor. Live: https://landfall-ohio.onrender.com Acces Code: 0725