CivicFix

Inspiration

We wanted an India where civic issues don’t get lost in the crowd—where citizens and authorities can collaborate to report, track, and fix problems transparently. CivicFix makes that possible.

What it does

CivicFix lets citizens snap photos of local civic issues, report them, and track them as NGOs or authorities resolve them. AI-powered analysis classifies issues, assesses severity, and predicts impact. Gamified karma points and leaderboards motivate engagement.

How we built it

Frontend:
- Mobile app: React Native + Expo with camera, map view, and leaderboard.
- Web app: Vite + HTML + CSS with feeds, maps, and leaderboard.
Backend: FastAPI gateway handling auth, CRUD, fix submission, and karma tracking.
AI Microservices: Google Gemini Vision for image analysis and multi-issue verification.
Search: Elasticsearch with hybrid kNN + geospatial + term filters.
Storage & Auth: Firebase for users/karma, GCS for images.
Infrastructure: Dockerized microservices for scalability and smooth deployment.

Challenges we ran into

Handling complex workflow of issue consisting of text and image uploads, processing, storage and retrieval.
Detection and fixing of issues using AI while minimizing hallucination.
Maintaining data consistency across Firebase, GCS, and backend services.
Designing system logic to avoid bottlenecks.

Accomplishments that we're proud of

Fully automated AI detection and verification pipeline.
Hybrid search combining kNN embeddings, geospatial filters, and term queries.
Scalable architecture supporting mobile and web clients.

What we learned

Complex AI integrations need careful batching and async handling.
Hybrid search + embeddings significantly improve retrieval accuracy.
Dockerized microservices simplify scaling and local development.
Teamwork, working on multiple fronts—app, web, and backend.

What's next for CivicFix

Allow social posting and tagging relevant authorities on social media platforms like X.
Allow deep analysis and comparison of fix data, issue upload etc.
Collect deep data like contact info—email, mobile nos etc.—of relevant departments for every block in India (for accurate and effective reporting), and probably expand to the world.