Inspiration
We’ve been frustrated by how absurdly hard it still is to report real infrastructure problems, especially in urgent places like school zones. 311 feels like sending issues into nothing. We wanted something that was instant, natural, and actually got things fixed, rather than logged and forgotten. Voice and AI finally made that possible.
What it does
You just speak. No app, no form, no typing(you can if you wish). “2-foot pothole outside Lincoln Elementary , cars are swerving into the bike lane.” CivicGrid immediately pulls location, understands risk, classifies the issue, and creates a work order with urgency and route to the right actor. No human backlog in between.
How we built it
We built a voice + image intake system that sends raw inputs into our Firestore pipeline. Claude → interprets the report → computes risk, priority, and geographic context → writes a normalized work item. From there, it’s ready for dispatch: to contractor and government approval. No manual triage. Designed from day one for real-world deployment.
Challenges we ran into
Mapping free natural speech to real-world geolocation is extremely hard. Avoiding false confidence is harder. We had to define an insanely thin yet expressive data model — minimal enough to move fast, structured enough to plug into city systems. Getting that balance right took iteration.
Accomplishments that we're proud of
The system works end-to-end: you speak, we generate dispatchable work order in seconds. We proved you can prioritize a school-zone hazard over random noise without human triage.
What we learned
Getting actual infrastructure to behave correctly is way harder than “just throw AI at it.” We understand government really tried hard on infrastructure design.
What's next for Civil Grid
Live school-zone pilot. Automatic contractor dispatch. Transparent public-facing “in-progress” tracker.
Log in or sign up for Devpost to join the conversation.