PantryPulse

Inspiration

Imbalance of donations is an issue food pantries face across the state. The effects of this can be observed first hand - in some pantries, over 20% of food is collected in surplus and rots, ultimately going to waste. Meanwhile 80% of food pantries report being unable to provide food to those they serve. Thus, we created a way for food pantries and donors to align their needs. With this solution, we can boost the sustainability and effectiveness of food pantries, and make sure all parties involved have access to the resources they need.

What it does

With PantryPulse, food pantries can communicate to an AI agent what foods they need and of what already have an abundance. Donors can consult the agent when making donations, learning which pantries are in need of the foods they are offering. If a pantry reports an excess of a certain food, the agent will direct those donations elsewhere. Alternatively, if a pantry is in need of a certain food, those donations will be directed towards it. Entities interact with the AI agent through SMS, which many users will already have. This keeps our software as accessible and easy-to-use as possible.

How we built it

We built PantryPulse around Photon as the main conversational and messaging layer. Photon receives incoming messages, manages the interaction with donors and pantries, and sends recommendations and follow-up messages through SMS. Behind the messaging layer, our backend uses pantry inventory data and a Pandas-based matching algorithm to determine where a donation is needed most. The algorithm compares each pantry's current inventory against its target inventory, ranks locations by shortage, and returns the highest-need pantry along with backup options. Photon then communicates those results back to the donor and can continue the conversation from there.

Challenges we ran into

One of our biggest challenges was keeping the system simple while still making useful decisions from real inventory data. We originally explored using a separate AI model for natural-language processing, but that introduced additional complexity and integration issues that were not necessary for our final prototype. We simplified the architecture by allowing Photon to handle the conversational logic while our backend focused entirely on deterministic pantry matching. We also had to account for cases where a pantry did not need a particular item, where multiple pantries had similar needs, and where no pantry currently reported a shortage.

Accomplishments that we're proud of

We're proud that PantryPulse turns a potentially complicated donation-matching process into something as simple as sending a text message. A donor does not have to download an app, search through pantry websites, or manually contact several organizations to determine where their food would be most useful. We also built a matching system based on real inventory shortages rather than arbitrary recommendations. By combining a lightweight messaging experience with a transparent matching algorithm, we created a prototype that is both easy to use and easy to understand.

What we learned

This project taught us a lot about simplifying system architecture and choosing the right tool for each part of a problem. Rather than adding AI components simply because they were available, we learned that a straightforward algorithm can often be more reliable for structured decisions such as ranking pantry needs. We also gained experience working with messaging APIs, Pandas, inventory data, Git collaboration, and the kinds of edge cases that appear when turning an idea into a working prototype under hackathon time constraints.

What's next for PantryPulse

Our next step is to make PantryPulse more aware of the real-world logistics involved in food donations. We would like to incorporate donor distance, pantry operating hours, expiration dates, storage capacity, and donation size when ranking locations. We also want pantries to be able to update their inventory directly through SMS so that recommendations stay current without requiring a separate application. Eventually, PantryPulse could automatically contact backup pantries when the first choice does not respond, coordinate donation confirmations, and help create a broader network in which surplus food is consistently redirected to the communities that need it most.

Built With

• photon
• python