TrashTalkers.tech

Fighting climate inaction through simplicity. Our app classifies waste, maps recycling spots, and shows monetary value turning confusion into environmental action and profit.

Comment 1

Inspiration

We were inspired by the growing environmental crisis and the disconnect between people's desire to recycle and their ability to do so effectively. Too many recyclable items end up in landfills simply because people don't know what can be recycled, where to take items, or if it's even worth the effort. We wanted to eliminate these friction points and empower users to make environmentally conscious decisions while potentially earning money through redemption programs. Our goal was to create a future where sustainability is effortless and accessible to everyone.

What it does

TrashTalkers.tech is an AI-powered waste classification platform that removes the guesswork from recycling. Users simply snap a photo of their waste item, and our Hugging Face powered image recognition instantly identifies and classifies it. The app then provides detailed disposal instructions generated by Google Gemini AI, estimates potential redemption values in the user's area, and displays nearby recycling centers on an interactive map with turn-by-turn directions. Each user has a personal dashboard to track their recycling history and environmental impact, making sustainable habits easy to maintain and measure.

How we built it

We built TrashTalkers.tech using Next.js 16 with the App Router for a modern, performance based frontend experience, styled with Tailwind CSS for a sleek, responsive interface. Firebase handles authentication and Firestore stores user data and item history. For the AI capabilities, we integrated Hugging Face's image classification models for waste identification and Google Gemini for generating detailed disposal guidance. The location features leverage Google Maps API and the new Google Places API to find nearby recycling centers and provide embedded directions. We containerized the entire application with Docker and set up a complete CI/CD pipeline using GitHub Actions to automatically deploy to Google Cloud Run, with secrets managed through Google Cloud Secret Manager.

Challenges we ran into

One of our biggest challenges was accurately classifying diverse waste items through image recognition. Garbage can come in countless forms, shapes and sizes. We had to fine-tune our Hugging Face model integration and implement confidence scoring to handle edge cases. Integrating multiple Google APIs (Maps, Places, Gemini) while managing API costs and rate limits required careful architecture decisions. The new Google Places API had limited documentation, forcing us to experiment extensively. We also struggled with making redemption value estimates accurate across different locations and item types, as these vary significantly by region. Finally, creating a seamless user experience that works across devices while handling real-time location services and map rendering required extensive testing and optimization.

Accomplishments that we're proud of

We're incredibly proud of creating a fully functional, production ready application that genuinely solves a real world problem. The AI classification system works remarkably well, providing accurate results with confidence scoring. Our automated deployment pipeline means we can iterate rapidly and push updates seamlessly. The user experience is smooth and intuitive so users can go from photo capture to finding their nearest recycling center in under 30 seconds. We successfully integrated three major APIs (Hugging Face, Google Gemini, Google Maps/Places) into a cohesive system that feels native and responsive. Most importantly, we built something that can actually motivate people to recycle more by removing barriers and adding incentives.

What we learned

This project taught us invaluable lessons about full stack development and API integration. We gained deep experience with Next.js 16's App Router and server side rendering capabilities, which significantly improved our application's performance. We learned how to effectively prompt and integrate large language models like Gemini for reliable, structured outputs. Working with Firebase Admin SDK taught us proper authentication patterns and server side security. We also learned that building for sustainability requires thinking beyond just functionality to consider motivation, habit formation, and reducing friction at every step.

What's next for TrashTalkers.tech

Our roadmap focuses on hyper localized recycling intelligence by building a comprehensive database of recycling rules across specific counties, cities, and states because what's recyclable in San Francisco may require special handling in rural Ohio. TrashTalkers will become the definitive local recycling guide with gamification, community leader boards, and calendar integration for special collection events, ensuring every user receives precise, actionable guidance tailored to their exact location.

Built With

Share this project:

Updates