Check. Connect. Collect.
| Societal Challenge | Partner Organization | AI Technology | Expected Impact |
|---|---|---|---|
| Small-store food surplus remains disconnected from food banks facing rising demand and costs. | Circle K (West Lafayette); Food Finders Food Bank (Lafayette). | Machine learning for surplus forecasts; optimization for pickup selection, routing and scheduling. | More food recovered, lower collection costs and stronger participation by small businesses. |
THE CHALLENGE
Near Lafayette, 86 of 92 food businesses had no publicly listed donation program, so their surplus rarely reaches a food bank. Small donations are hard to recover: availability is uncertain, pickups are costly, and stores have little incentive to join. Food banks need a practical way to turn this scattered surplus into dependable supply.
OUR AI SOLUTION
FreshGuard combines a food-bank planning dashboard with an engineering-based toolkit for small businesses.
The dashboard brings together surplus forecasts, existing routes and receiving capacity to recommend which donations to collect, when to collect them and how to route pickups. AI is needed because each store’s surplus is small and uncertain: forecasts estimate availability, and optimization picks the pickups worth making within capacity and time limits. This has been partially validated in the Phase 1 prototype.
The toolkit helps owners develop their own custom software using AI based on our prototype. It guides them through defining requirements, building tools, testing outputs against operating data and correcting errors, with industrial engineering methods for verification.
Together, these tools address both sides of participation: making donations worthwhile to collect and offering stores practical technology guidance that can reduce waste and improve operations. The pilot will test whether this benefit attracts and retains donors.
PROGRESS SINCE PHASE 1
| Phase 1 | Phase 2 Progress |
|---|---|
| Inventory inference model and store-checking prototype built on Circle K data. | Dashboard prototype for surplus collection and pickup planning. |
| Simulated at one store: 40% less inspection time, 20% less overstocking waste, about 670 lb of surplus a year. | 10-store pilot and evaluation plan; software toolkit as a participation incentive to test. |
Evidence status: Current efficiency and collection-cost estimates are modeled results. Phase 2 will test performance in operating stores and real pickups.
WHY IT MATTERS
PEOPLE
People facing food insecurity could receive more locally recovered food. Food banks gain another source of supply, while small businesses gain support to build reliable tools tailored to their operations.
IMPACT
Recover donations that would otherwise be missed, make small pickups more economical, and establish a repeatable way for independent businesses to join food-bank networks.
WORKING WITH PARTNERS
Circle K (1209 Sagamore Pkwy W, West Lafayette)
Our Circle K partner provided sales data and insight into stocking practices. The store lacks item-level tracking and relies on manual checks, which grounded the Phase 1 models in actual operations. In Phase 2, we plan to compare forecasts with physical checks and gather staff feedback on the donation workflow. These findings will refine the models and toolkit to fit small-business constraints.
Food Finders Food Bank (Lafayette)
Consultation with Food Finders informs how FreshGuard could fit foodbank operations. Subject to a formal partnership, we propose reviewing donation eligibility, handling requirements, pickup coordination and receiving capacity together. This collaboration would help us test whether dashboard recommendations are practical to carry out and what must be in place before expanding through food-bank networks.
PATH TO SCALE
Request: $50,000 over 12 months, in two six-month stages.
Months 1–6 | $25,000: Pilot and validation
- Recruit 10 stores; track enrollment, retention and donation frequency.
- Build the dashboard and toolkit; test participation incentives.
- Compare forecasts with physical checks; measure food recovered, pickup completion and cost per pound.
Months 7–12 | $25,000: Repeatability and sustainability
- Test onboarding with less hands-on support, and whether owners can build and verify tools with the toolkit.
- Document onboarding, ownership, maintenance and recurring costs. After the grant, food banks keep the dashboard if it lowers cost per pound; each food bank and its nearby stores is one replication unit.
PHASE 2 GOALS
- Establish an active donor pilot: recruit 10 stores and track enrollment, retention and donation frequency.
- Validate collection performance: measure forecast accuracy, food recovered, pickup completion and actual cost per pound.
- Test the software-development toolkit: assess whether owners can develop and verify useful tools, and record the support required.
- Deliver a sustainable operating plan by month 12: document onboarding procedures, maintenance responsibilities and recurring costs.
THE TEAM
- Student leads: Mark Chen and Hazel He
- Advisor: Dr. Reem Khir (Website)
- Edwardson School of Industrial Engineering, Purdue University