ShelfLife

Inspiration

About 40% of food in the United States gets thrown into landfills by individuals and retailers while families in the United States increasingly face food insecurity. The recent removal of SNAP for low-income households exacerbates this risk of food insecurity may increase people’s consumption of expired produce. In addition to the food loss and waste (FLW), the inappropriate disposal of food waste contributes to Greenhouse Gas emissions globally by adding methanous gases into the atmosphere that increases global warming and climate change.

What it does

ShelfLife encourages households to eat healthier meals and cut down food waste by keeping track of the freshness of purchased produce, suggesting appropriate storage for each produce, and sending reminders to users on the expiration of their produce. It includes recipe suggestions from Chef Who, an AI chatbot chef that uses this inventory, datetime location and user preferences to create recipes, encourage community sharing of cooked meals for bulk items losing freshness, and suggest disposal methods that have the least impact on the environment.

How we built it

ShelfLife was developed as a full-stack web application using TypeScript for the frontend and Go (Golang) for the backend, deployed through Vercel for rapid, serverless scalability. The system architecture combines AI-driven decision support with real-time data management to help users track, predict, and act on food freshness.

We implemented a visual AI model to recognize food items and estimate their shelf life from user-submitted photos or barcode data. This predictive layer interprets image features and environmental cues to approximate spoilage timelines. Each item’s state is stored in a PostgreSQL database, with freshness automatically updated through Supabase triggers and Edge Functions that handle expiration logic and push notifications.

The frontend was designed for intuitive interaction — users can visualize their pantry, receive alerts for expiring items, and connect directly with Chef Who, our AI-powered recipe recommender. Chef Who uses OpenRouter for multi-model language generation, producing creative, waste-reducing meal suggestions based on what users already have or a sustainable disposal method if expired.

For background tasks, we employed AWS cron jobs to schedule freshness checks and notify users proactively. The project’s infrastructure is virtualized in our AWS cloud environment, which hosts the backend, frontend, and evolving AI agent services.

Continuous Integration and Deployment were set up via GitHub Actions, ensuring fast iteration cycles, automated testing, and smooth delivery across development stages. Together, this stack enables a scalable, energy-efficient platform capable of growing from household pilots to enterprise food-waste management systems.

Challenges we ran into

We faced integration issues when linking the AI recognition module with the backend database, and optimizing API calls for real-time freshness updates.

Accomplishments that we're proud of

We successfully built a cross-platform prototype capable of syncing household inventory, predicting spoilage with confidence intervals, and linking the results to a conversational assistant (Chef Who). Our model achieved stable real-time updates without data loss during concurrent sessions.

What we learned

We learned how to integrate AI, sustainability, and behavioral nudging into a unified system that transforms data-driven insights into everyday food decisions. This process deepened our understanding of backend optimization, prompt engineering, and user-centered environmental design. Most importantly, it highlighted how small, intentional household actions can scale into measurable contributions toward global sustainability goals.

What's Next

Scale for impact: Extend ShelfLife to support small businesses, restaurants, and retailers in managing inventory, optimizing procurement, and reducing commercial food waste.
Build community: Develop the community-sharing and redistribution features to enable surplus exchange between households, local organizations, and food banks.
Advance prediction: Incorporate environmental data (temperature, humidity, and seasonality) to improve spoilage forecasts and enhance the precision of AI-driven recommendations.