meri dharani

• 

Inspiration

India generates over 62 million tons of waste every year, yet less than half is processed scientifically. The rest piles up on streets, open lands, and water bodies — not because solutions don’t exist, but because coordination between citizens, workers, and governments is broken.

We noticed a common frustration everywhere:

Citizens want cleaner neighborhoods but don’t know where or how to report waste effectively

Sanitation workers receive inefficient, random assignments

Officials lack real-time visibility and are forced to react instead of prevent

After studying waste systems in Indian metro cities and global leaders like Finland, Singapore, and South Korea, we realized the missing piece wasn’t hardware — it was an intelligent, unified coordination platform. That insight became Meri Dharani.

What it does

Meri Dharani is an AI-powered smart waste management ecosystem that connects citizens, cleanup workers, and government officials on one real-time platform.

Citizens report waste with photos

AI validates and classifies waste automatically

Workers receive optimized cleanup routes

Officials get live dashboards, heatmaps, and predictions

The platform transforms waste management from complaint-driven to data-driven, ensuring accountability, transparency, and sustained community engagement.

How we built it AI & Machine Learning

Llama 3.2 Vision (11B) for real-time waste detection and classification from photos

Llama 3.3 (70B) multilingual AI chatbot via Groq Cloud (Hindi, Telugu, English)

Predictive models for waste hotspot forecasting

Backend

FastAPI for high-performance APIs

AWS Lambda for scalable, serverless notifications

MongoDB Atlas for real-time data storage

Google OAuth for secure authentication

Frontend

Progressive Web App (PWA) with offline support

Mapbox for interactive live maps and heatmaps

Tailwind CSS for a clean, responsive UI

Full multilingual support

Challenges we ran into

Accurate waste detection: Differentiating real waste from irrelevant images

False reporting: Preventing misuse while keeping the system accessible

Route optimization: Balancing efficiency with worker flexibility

Scalability: Designing a system that works from a single ward to an entire state

Stakeholder alignment: Building one platform that serves citizens, workers, and officials equally

Accomplishments that we're proud of

Built a complete end-to-end AI-driven waste coordination system

Achieved 95%+ accuracy in waste classification using Vision AI

Designed a verified cleanup system using before/after images

Created a fair gig-economy model for sanitation workers

Enabled real-time heatmaps and predictive analytics for governments

Delivered a fully functional PWA, not just a concept

What we learned

Waste management is more about behavior and coordination than technology alone

Citizens engage more when they see real impact and rewards

AI can dramatically reduce manual workload for governments

Transparency builds trust across all stakeholders

Designing for inclusivity (language, accessibility, offline use) is critical in India

What's next for Meri Dharani

Pilot launch in Andhra Pradesh with municipal partnerships

Integrate carbon credit tracking and ESG reporting

Expand AI predictions for seasonal and event-based waste surges

Add smart bin & IoT integrations

Scale to 4,000+ urban local bodies across India

Adapt the platform for other developing nations facing similar challenges

Built With

• ai-&-machine-learning-llama-3.2-vision-(11b)-for-real-time-waste-detection-and-classification-from-photos-llama-3.3-(70b)-multilingual-ai-chatbot-via-groq-cloud-(hindi
• challenges
• english)-predictive-models-for-waste-hotspot-forecasting-backend-fastapi-for-high-performance-apis-aws-lambda-for-scalable
• full
• multilingual
• ran
• responsive
• serverless-notifications-mongodb-atlas-for-real-time-data-storage-google-oauth-for-secure-authentication-frontend-progressive-web-app-(pwa)-with-offline-support-mapbox-for-interactive-live-maps-and-heatmaps-tailwind-css-for-a-clean
• support
• telugu
• ui
• we