Savr | Devpost

Cornell University generates over 700 pounds of waste per person annually, with food waste being a huge part of that problem. We noticed a clear disconnect: students throw away expiring groceries they forget about in their dorms, while dining halls like Morrison, Okenshields, and Becker throw out perfectly good food every single day. The real issue is that individual student waste and institutional dining waste happen side by side with zero connection between them. On top of that, some students on campus are dealing with food insecurity while food that could feed them is being tossed. We built Savr to solve this problem and make food waste prevention something students do every day without thinking twice about it. Savr is an AI-powered app that helps Cornell students prevent food waste, save money, and actually connect with their community. Smart Fridge Inventory lets you track what you have and when it expires with a simple interface that highlights what's about to go bad. The AI Recipe Generator uses Claude to look at what you actually have and create personalized recipes tailored to your skill level and equipment. Meal Planning generates 3-day meal plans that use ingredients smartly by prioritizing what needs to get eaten first. The Surplus Sharing Network connects you directly with dining halls and local restaurants in real-time so you can grab surplus food before it gets thrown away. We built this using React, Tailwind CSS, and Lucide Icons on the frontend, with Claude Sonnet 4 API powering the AI. Users can add items manually through a searchable dropdown or by taking a photo of their fridge for automatic identification using Claude Vision. The biggest challenge we faced was getting computer vision to actually work. The fridge image recognition kept misidentifying items and always returned "apples and milk" no matter what you photographed, so we improved the Claude Vision prompts with better guidelines and added a confirmation popup where users can verify and edit identified items. We also had issues with bulk item storage where only one item got saved to the fridge when multiple items were identified, so we refactored the backend to batch-process all identified items. The meal plan feature wasn't generating recipes based on actual fridge inventory, so we rebuilt the algorithm to query current contents first before sending to Claude. The surplus food tab was stuck showing the same "leftover bagels from Trillium" repeatedly, so we expanded keyword tracking to 50+ common food surplus terms and integrated multiple dining halls plus local restaurant partners to make it feel authentic. Despite these obstacles, we built a fully functional app that actually makes food waste prevention something students want to use every day. The recipe generator produces genuinely creative meal suggestions from whatever ingredients you have and feels personalized rather than generic. The waste prevention counter shows users exactly how many pounds of food they've saved from the landfill, creating a real connection between daily actions and actual impact. Looking forward, we want to partner with Cornell Dining Services so they can see anonymized data on what surplus items get claimed fastest, helping them reduce waste before it happens. We'll integrate with Cornell's meal plan system so you see dining hall menus in-app and get recipe suggestions that complement what's being served. We're exploring barcode scanning for faster item entry and working with local food banks to ensure unclaimed surplus food gets to students who need it. We'll add dorm-level waste prevention competitions with actual incentives and build an API so student sustainability organizations can use Savr.