YouTube Demo

Empowering Community-driven Environmental Actions

verde (“green” in several Romance languages) is an app providing a real-time, interactive map that serves as a hub for community-led environmental action.

This map is managed and enriched by agents, which asynchronously pull data, cluster issues, and organize environmental events.

This project was built in under 24 hours as a submission to the 2025 HackMIT global hackathon.

Inspiration

Have you ever noticed an environmental issue but not been able to address it alone? 🤔

Our current model of environmental action just can’t keep up; cities receive hundreds of thousands of environmental complaints per year, but community efforts remain limited. That’s because of the current top-down framework for environmentalism, where people rely on a few large organizations to plan events, which is slow, bureaucratic, and disconnected from the local issues people see in their own neighborhoods every day.

We saw an opportunity to flip this model on its head with the power of multimodal ai agents and crowdsourced data. What if we could empower individuals to identify problems and organize their own solutions? ✨

Features

📸 Report Anything: Any user can identify an environmental issue (e.g. an overflowing trash can) and drop a “Issue” marker on the map by simply taking a picture of the scene. Our system automatically identifies and verifies the environmental issues present in the image.

🤝 Organize Whenever: Other users can see these reported issues and vote on the credibility and update real-time whether the issue remains. If an issue is easy to resolve, users can simply arrive at the location indicated by the pin and help resolve it. If an issue is difficult to resolve and requires more volunteers, anyone (an individual, a local club, or an NGO) can then create an “Event” marker linked to an issue, such as scheduling a park cleanup for the upcoming weekend.

🤖 Agentic Enrichment: Our system frequently fetches the most up-to-date government and NGO data and extracts information about the most pressing local environmental issues in each community. It then clusters these issues by neighborhood and automatically creates and organizes community-driven events centered around resolving these issues.

🏆 Gamified Element: The community can RSVP to events, coordinate through the platform, and take action. To motivate participation, we’ve incorporated a gamification system where users earn points for reporting issues, creating successful events, and participating in them.

Technology

We built this project with a focus on cross-platform accessibility, enabling the democratization of environmental actions for all users on all devices. 📱 Frontend (Mobile App): The application is built using Flutter. This allowed us to write a single codebase that compiles natively for iOS, Android, and the web, ensuring a consistent user experience.

💾 Backend & Database: We chose Supabase as our backend-as-a-service. This saved us immense development time. We used Supabase Functions to handle server-side logic, such as awarding points and managing marker data.

🧠 AI & Data Processing: We implemented a powerful but environment-friendly solution for issue verification. Using Anthropic’s Claude models, we were able to accurately identify environmental issues and generate concise, natural, and convincing descriptions from messy, multimodal data. By using Anthropic’s powerful but energy-efficient language models, we also avoid the cost of training domain-specific, large neural networks for classification and credibility judging.

🚀 Dev Tools: A large portion of the implementation details are generated by Windsurf. This allows us to focus on brainstorming ideas, designing, and integrating multi-sourced data and LLMs instead of writing tedious, repetitive code.