Why we built it
In a fast fire, the evacuation order is a broadcast, but rescue is one door at a time. The people who can't leave on their own depend on a few lift vans and ambulance crews, and dispatch usually learns who needs them from the phone lines, while thousands of others are calling too. Registries exist, but calling them one by one reaches only some people in time.
We asked a different question: what if the first thing that happened at SET was a phone call to every single home, at the same time, and the answers went straight into a plan?
What it does
An AI voice agent calls every household in the zone in parallel and builds a census: who needs a ride and what kind, who can drive and has a free seat, who already left. A planner matches riders to vehicles every few seconds, puts wheelchair and stretcher riders first, and keeps seats honest trip by trip. Drivers accept on their phones; only then is the resident called back with the driver's name and ETA. The console shows the whole operation on one map, and Play evacuation replays the entire night.
What it won't do
It won't tell anyone help is coming until a driver has accepted. Ride status, ETAs and driver names come from fixed templates filled from the database, never from the language model, and every line passes an honesty check. It doesn't replace 911 or the fire department; it gives them a list they can act on.
Data and simulation
The homes, people and fire on these pages are simulated and marked SIM. The North Scottsdale world has 6,000 homes whose addresses follow the county's address grid; the fire is a model run, not a forecast. The share of people who need help, their mobility needs and the timing assumptions come from public sources: U.S. Census ACS tables for the Oak Creek Canyon and Sedona areas, the official record of the June 2026 Pocket Fire, and published figures from the 2025 Los Angeles fires and Hurricane Beryl. When you use Try it, your phone number and the address you type are used for that demo call, its callbacks and your volunteer car.
Team
Four people, 24 hours, SunHacks 9.0 at Arizona State University, November 2026.
- PRPremPlatform, infrastructure and integration
- YCYashResponder side: console, driver app, dispatch, fire map
- DPDeepResident side: voice agent, calls, honesty checks
- ABAaryaStory, data, research with responders, video