Flare Kingston

Inspiration

Flare Kingston was inspired by how difficult it still is for cities to efficiently process resident reported issues. Problems like broken streetlights, unsafe sidewalks, or damaged roads often rely on manual review, which slows response times and buries urgent cases. Living in Kingston, we wanted to explore how AI and cloud infrastructure could make civic reporting faster, clearer, and more actionable for city staff.

What it does

Flare Kingston is an AI powered civic reporting platform where residents submit a photo, short description, and location of an issue. The system automatically classifies the problem, detects duplicates, estimates urgency, and generates plain language summaries that help city employees understand what matters most. Instead of scrolling through raw reports, staff see a ranked and clustered view of issues that highlights safety risks and emerging patterns.

How we built it

The platform is built entirely on AWS using a serverless architecture. Images are uploaded to Amazon S3 and trigger AWS Lambda functions for processing. Amazon Rekognition and AWS Bedrock analyze images and text to label issues, assess urgency, and generate summaries. Reports are stored in DynamoDB, where AI generated priority scores allow issues to be sorted, clustered, and tracked. A lightweight frontend allows users to submit reports and city staff to view insights in real time.

Challenges we ran into

One of the main challenges was designing an AI pipeline that felt realistic and trustworthy rather than experimental. We learned how to build event driven systems, structure NoSQL data for real world queries, and use AI not just for classification but for decision support. The biggest takeaway was how powerful cloud native tools can be when combined with a clear real world problem.

Accomplishments that we're proud of

We built a fully functional, end to end civic reporting pipeline that goes from a resident submitted photo to an AI prioritized issue in a city dashboard. The system successfully classifies reports, detects duplicates, ranks urgency, and generates plain language summaries using a serverless AWS architecture. We are especially proud that the solution feels realistic, scalable, and deployable rather than purely conceptual.

What we learned

Through this project, we learned how to design event driven systems using AWS Lambda and S3, structure NoSQL data in DynamoDB for real world queries, and apply AI models in a responsible and practical way. We also learned how to balance technical ambition with clarity, making sure the AI output is understandable and useful for non technical users like city staff.

What's next for Flare Kingston

Next, we want to expand the platform with a richer analytics layer that highlights trends over time and across neighborhoods. We also plan to improve the city dashboard with filtering, map based views, and department specific workflows. Long term, we see Flare Kingston evolving into a deployable civic tool that can be adapted for other cities, campuses, and large events.

Built With

• amazon
• amazon-web-services
• api
• bedrock
• dynamodb
• gateway
• lambda
• react
• rekognition
• s3
• sns