Project Info

UrbanResponse AI

Devpost

This project did not submit a demo video on Devpost.

Inspiration

Emergency response coordination is a critical challenge that affects every community. During the Southern California wildfires earlier this year and recent campus emergencies at UC Berkeley, we witnessed firsthand how fragmented communication and slow decision-making can cost precious time when lives are at stake. We were inspired by the potential to combine cutting-edge AI with immersive 3D visualization to transform how cities respond to emergencies. Our team envisioned a future where AI agents could instantly analyze disaster scenarios, optimize resource allocation, and coordinate multi-agency responses; all while emergency managers watch it unfold in stunning photorealistic 3D.

What it does

AI-Powered Decision Making: Google Gemini LLM analyzes emergency events in real-time, considering severity, location, resource availability, and Berkeley-specific factors to make intelligent dispatch decisions. Immersive 3D Visualization: Using Google to render Berkeley's buildings and terrain with stunning detail, allowing emergency managers to see exactly where events occur and track vehicle movements in real-time. Multi-Agent Coordination: CrewAI agents work as a specialized team - a Triage Agent assesses threats, Resource Allocation Agent optimizes vehicle deployment, and Dispatch Agent coordinates responses. Hardware Integration: ESP32 touch-screen interfaces deployed throughout Berkeley allow field personnel to trigger emergencies instantly, bridging the physical-digital gap. Smart Vehicle Dispatch: The system automatically selects optimal emergency vehicles (ambulances, fire trucks, police) and calculates realistic routes using Google Maps' emergency routing.

How we built it

Frontend Architecture: JavaScript ES6+ with Google Maps JavaScript API v3.55 Google 3D Maps API (maps3d) for photorealistic visualization AI Backend: Python FastAPI serving the core intelligence system Google Gemini 2.0 Flash as the primary LLM for emergency analysis CrewAI v0.28.8 for multi-agent coordination and workflow Fetch.ai uAgents for decentralized agent communication Hardware Layer: ESP32 microcontrollers with 4.0" TFT touch displays

Challenges we ran into

One issue we faced early on was synchronizing Google Maps routing with vehicle movement animations while maintaining realistic speeds. Handling route calculation failures also proved technically challenging. This was a core functionality for simulating emergency responses in real time. Similarly, ensuring that the Gemini LLM provided consistent and logical emergency responses across different event types, severities, and sources required extensive prompt engineering and testing. Finally, balancing CrewAI's powerful capabilities with real-time performance requirements meant developing both full and simplified agent versions that could adapt dynamically to system load.

Accomplishments we're proud of

Given it was our first time using Agentic AI, we were able to adapt quickly while exposing ourselves to new technologies. We persevered through numerous challenges and created a system that has a real-world impact on a lot of people. In addition, we achieved sub-5-second emergency event processing from hardware trigger to AI analysis to vehicle dispatch visualization, which simulates the real-life effectiveness.

What we learned

Through this project, we learned how to bring together diverse technologies, AI, real-time simulation, embedded hardware, and geospatial data, into a unified, mission-critical system. We developed a deeper appreciation for designing user-centric tools that are both intelligent and responsive under pressure. Working with large language models and agent-based systems taught us the importance of adaptability, prompt design, and clear coordination. Above all, we learned how to turn a bold idea into a tangible, impactful solution through collaboration, iteration, and resilience.

What's next

Multi-City Expansion: Extend beyond Berkeley to San Francisco, Oakland, and eventually any city with emergency response needs. Hospital Integration: Connect with Berkeley's hospital systems for real-time bed availability and specialized care routing. Public Alert System: Add civilian notification capabilities with evacuation routing and safety instructions. Multi-Agent Decision Making Process

Analysis

Compare with all teams

View

Metric

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only
  • AnthropicIn code
  • C++In code
  • CrewAIIn code
  • CSSIn code
  • FastAPIIn code
  • HTMLIn code
  • JavaScriptIn code
  • PythonIn code
  • Google GeminiClaimed

8 of 9 appear in the indexed code. 1 claimed on Devpost could not be matched to code, which may simply mean the tool leaves no trace in the repository.

AI coding agents

  • GitHub CopilotCommits

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

565 KB

Source files

27

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.

0 stars