LemonLens is an operational analytics platform built for the Lemontree nonprofit as part of the Morgan Stanley Code to Give Hackathon.
Built with the same tech stack as Lemontree (Next.js, React, TypeScript, Vercel), LemonLens is designed for seamless integration into the existing platform.
The system transforms community feedback, pantry reviews, and food inventory photos into structured operational insights that help food banks, donors, and partners make faster and more informed decisions about where food resources are needed most.
Instead of manually reviewing scattered feedback and images, LemonLens automatically categorizes food supply data, analyzes service signals such as wait times and unmet demand, and surfaces priority insights through an interactive dashboard.
Lemontree has rich data on 1,400+ food pantries, but partners like the NYC Mayor's Office can't easily explore it.
Lemontree collects valuable community feedback about food access across New York City, including pantry reviews, wait times, and photos of available food resources. However, this information is difficult for partners to organize, interpret, and act on in real time.
Current challenges include:
- Pantry feedback and photos are unstructured and require manual review
- No way to filter pantries by dietary needs (Halal, Kosher, Protein, Dairy, etc)
- Operational issues such as shortages or long wait times are difficult to detect quickly
- Data is difficult to aggregate across neighborhoods or food categories
- Demographic context (poverty rates, SNAP enrollment) isn't connected to pantry data
- Partners lack a simple way to identify which pantries require urgent support
As a result, food banks and donors often rely on incomplete or delayed information when deciding where to send resources.
LemonLens converts community feedback into real-time operational intelligence.
The system automatically processes pantry photos, categorizes food inventory, analyzes demand signals, and visualizes trends through an interactive dashboard.
This allows partners to quickly identify shortages, service disruptions, and high-demand locations across the city.
Instead of manually interpreting reviews and images, partners can immediately see:
- Which pantries are under the most pressure
- What types of food are currently missing
- Where operational issues are emerging
- Where the next shipment of resources should be sent
Clients or volunteers visit a pantry and submit a mobile review.
Data captured includes:
- Wait time in minutes
- Whether assistance was received
- A photo of the available food
- Optional written feedback
This creates a real-time signal about pantry operations and available resources.
The system demonstrates an AI-powered classification pipeline using Claude Vision API architecture.
Pantry photos are analyzed and structured food tags are extracted:
- Produce
- Protein
- Dairy
- Grains
- Canned goods
- Halal / Kosher indicators
This replaces hours of manual photo review and converts visual data into structured inventory signals that power the dietary preference filters.
Extracted food tags are standardized into supply profiles.
Raw detections are grouped into consistent food categories that support filtering, aggregation, and analytics across locations.
For example, item names like:
- "Campbell's Chicken Noodle Soup"
- "Jack & the Beanstalk Long Grain White Rice"
- "Cheese Club Macaroni and Cheese"
need to be standardized before they can support product-level insights. This helps group messy item detections into shared food categories, combine pantry supply data with other datasets such as demographics, and power dashboard filters and category-level analytics
LemonLens generates a real-time Needs Score for each pantry.
This score combines:
- Detected food supply
- Wait time signals
- Reports of unmet demand
- Operational feedback from users
The result is a priority ranking that highlights which locations require immediate support.
In addition to pantry supply signals, the system supports a lightweight feedback pipeline for capturing operational experience data in structured JSON.
This includes fields aligned with Lemontree’s review model, such as:
- whether the client got help
- rating
- written feedback
- wait time
- whether listing information was accurate
- reasons a client could not receive help
This feedback layer is important because it complements pantry inventory data with real user experience signals.
Partners interact with the data through a web dashboard that provides:
A Mapbox-powered visualization of more than 1,400 food resource locations across New York City.
Locations are clustered dynamically and update based on viewport and filters.
Users can filter resources by:
- Borough
- Food category (Fresh Produce, Dairy, Meat, Grains, Canned)
- Dietary preferences (Halal, Kosher)
- Service type
- Operational priority
This enables targeted exploration of supply gaps and service demand, including finding pantries that match cultural dietary needs.
A live ranking of the highest-need pantries based on the Needs Score.
This allows organizations to quickly identify where food shipments or operational support should be directed.
The platform generates operational summaries including:
- City-wide food supply breakdowns
- Category shortages
- Demand hotspots
- Service disruptions
Partners can export shareable PDF reports to coordinate with donors, logistics teams, and community organizations.
Filter pantries by food availability including Fresh Produce, Dairy, Meat, Grains, Canned Goods, Halal, and Kosher options. Color-coded tags on each pantry card show what's available at a glance.
Automatically detects food categories from pantry photos using computer vision.
Ranks pantries by urgency using real-time operational signals.
Mapbox-powered visualization of 1,400+ NYC food locations with clustering and viewport filtering.
Dynamic charts showing distribution of food categories across the city.
Aggregates user feedback including wait times, service success, listing accuracy, and unmet demand signals to identify recurring operational issues.
Visualize food access alongside real Census (ACS) and USDA data including poverty rates, SNAP enrollment, food desert indicators, and vulnerable population density.
LemonLens helps partners answer a critical operational question:
Where should the next shipment of food go right now?
By transforming raw community feedback into structured analytics, the platform enables faster and more informed decisions about food distribution and resource allocation.
This helps improve service reliability, reduce shortages, and strengthen food access across communities.
Potential extensions of the platform include:
- Predictive models for food demand forecasting
- Automated alerting for emerging shortages
- Mobile-optimized volunteer check-in flow
- Integration with donor inventory systems
-
Clone the repository git clone https://github.com/your-repo/lemonlens.git cd lemonlens
-
Install dependencies npm install
-
Configure Environment Variables Create a .env.local file in the root directory:
- ANTHROPIC_API_KEY=your_api_key
- NEXT_PUBLIC_MAPBOX_ACCESS_TOKEN=your_mapbox_token
-
Run the development server npm run dev
LemonLens was developed during the Morgan Stanley Code to Give Hackathon for Lemontree NYC.
The project demonstrates how data processing and visualization can transform community feedback into actionable insights that strengthen food access operations.
Team Members:
- Ishrat Arshad
- Rohit Karnik
- Anish Yenduri
- Nirmit Bhoyar
- Philip Shaji Baby
Acknowledgments
We would like to thank Morgan Stanley for hosting the Code to Give Hackathon and providing this platform for social impact. We also express our sincere gratitude to our mentor, Nirali Maniar, for her invaluable guidance, support, and technical feedback throughout this project!
This project is open source and available under the MIT License.