Building HeatSentry : Multi-Agent Urban Heat Resilience with Google Gemini.
*Disclaimer: I created this piece of content for the purpose of entering the Google #AllThingsAgentic Hackathon .
Extreme heat is among the deadliest climate hazards facing cities. In Phoenix, asphalt temperatures frequently exceed 140°F (60°C).
To tackle this, I built HeatSentry—an autonomous multi-agent command system powered by Google Gemini models and grounded in FortyGuard’s 2-meter pedestrian thermal mesh API.
What It Does
HeatSentry automates municipal heat response through 10 collaborative AI agents that monitor, negotiate, and dispatch physical interventions in real time:
- Cool Pavement Routing: Dispatches titanium-dioxide reflective coating trucks to high-heat corridors.
- Mobile Misting Fleets: Deploys high-pressure evaporative cooling trailers to transit hubs.
- Grid Peak Shedding: Curtails commercial HVAC loads during thermal peaks.
- OSHA Labour Shield: Dynamically enforces mandatory work/rest cycles for outdoor workers based on Wet-Bulb Globe Temperature (WBGT).
- Vulnerable Citizen Sentinel: Directs emergency hydration and outreach to high-risk zones.
Google AI & Agentic Stack
- 10-Agent Negotiation Mesh: Uses game-theoretic Nash bargaining consensus with a deterministic supervisor circuit breaker to resolve resource conflicts.
- Google Gemini 3.7 Flash: Powers the real-time Incident Copilot via Server-Sent Events (SSE) token streaming.
- Google Gemini 3.1 Flash Audio: Synthesizes instant 24kHz neural voice emergency alerts in English, Spanish, and Arabic.
-
Google Gemma (
gemma-2-9b-it): Delivers on-device field triage with sub-20ms edge response. - SHA-256 Block Ledger: Immutably records every dispatch decision for FEMA Category B audit compliance.
Links
Live Application
https://ais-pre-ul2mlgbobcxgg7tkb2lkrl-305446503352.asia-southeast1.run.app
#AllThingsAgenticHackathon #GoogleAI #ClimateTech #AI
Top comments (0)