Inspiration LumenRoute AI was inspired by the dual challenges Georgia commuters face: traffic safety and environmental impact. We wanted to move beyond passive GPS routing by turning existing GDOT infrastructure into a proactive "Guardian" system that understands the road as well as a human does.
What it does LumenRoute AI analyzes live feeds from over 50 GDOT traffic cameras in real-time using Gemini 1.5. It detects hazards like flooding and accidents, suggests "Eco-Routes" optimized for lower carbon emissions, and provides a "Scout Mode" for community-driven infrastructure reporting.
How we built it We built a full-stack application using React, Vite, and TypeScript for the frontend, with Leaflet for mapping. The backend is powered by Bun and Hono, chosen for high-performance edge capabilities. We integrated Google Gemini 1.5 Flash for vision analysis and Gemini Pro for reasoning, all containerized with Docker and deployed on DigitalOcean.
Challenges we ran into Our biggest hurdles were handling the massive data throughput of GDOT's ArcGIS MapServer and configuring Docker networking for Nginx to correctly route SPA traffic. We solved these by implementing a backend "diffing" strategy for data polling and writing custom Nginx logic to handle deep linking.
Accomplishments that we're proud of We are particularly proud of our "Scout Mode" verification system. Using Gemini Vision to instantly verify if a user-submitted photo actually shows a road hazard (like a pothole) allows the system to be self-moderating and spam-resistant.
What we learned We learned how to manage real-time multimodal data streams—combining visual camera feeds, text-based reports, and environmental sensor data—to make complex, automated routing decisions.
What's next for LumenRoute AI We plan to expand our analysis to include predictive traffic modeling using historical GDOT data and integrate more granular air quality sensors to refine our Eco-Route carbon emission calculations.
Log in or sign up for Devpost to join the conversation.