Zarfo – AI Powered Food Waste Reduction Platform

Zarfo is an AI-powered platform that reduces food waste from hotels and restaurants by deciding whether surplus food should be sold at a discounted price or donated to NGOs and shelters before expiry. The system combines Machine Learning, dynamic pricing, and food redistribution to ensure surplus food is utilized efficiently.

Problem Statement

Restaurants generate large amounts of surplus food daily which often goes to waste due to:

• Lack of demand prediction
• No pricing strategy for near-expiry food
• Poor coordination with donation networks

Zarfo solves this using AI-based decision making and automated redistribution.

Key Features

AI Decision System

Predicts whether food should be:

• SELL (discounted price)
• DONATE (to NGOs or shelters) based on expiry time, demand probability, quantity, and price.

Dynamic Pricing

Food prices automatically decrease as expiry approaches to increase chances of selling.

Time Left	Price
3 hours	₹100
2 hours	₹80
1 hour	₹50

Food Lifecycle Monitoring

An AI agent continuously monitors items and updates decisions: SELL → PRICE REDUCTION → DONATE

Donation Network

Unsold food is redirected to:

• NGOs
• Orphanages
• Old-age homes
• Food shelters

System Architecture

Frontend (React + Vite)
        |
Node.js Express Backend
        |
FastAPI ML Service
        |
MongoDB Database

Tech Stack

• Frontend: React, Vite, Tailwind CSS
• Backend: Node.js, Express.js
• AI / ML: Python, FastAPI, Scikit-learn, RandomForest Pandas, NumPy
• Database: MongoDB

Machine Learning Models

Food Decision Model (Classification)

Predicts whether food should be SELL or DONATE.

Inputs:

• Shelf Life
• Time Left
• Quantity
• Original Price
• Food Category
• Demand Probability

Output:

SELL or DONATE

Dynamic Price Model (Regression)

If the item is SELL, the model predicts the optimal discounted price.

Inputs:

• Time Left
• Shelf Life
• Demand Probability
• Quantity
• Original Price

Output:

Suggested Discounted Price

Dynamic Pricing Logic

life_ratio = time_left / total_shelf_life
new_price = original_price * life_ratio

Business Rule

Certain rules override ML predictions:

If time_left <= 2.5 hours:
decision = DONATE

Future Improvements

• Route optimization for food donation
• Demand prediction using historical sales
• Reinforcement learning for pricing
• Mobile app for NGOs and volunteers
• Google Maps API integration

Installation

Clone repository:

git clone https://github.com/farhat-1203/zarfo.git

Backend:

cd backend
npm install
npm start

ML Service:

cd models
pip install -r requirements.txt
uvicorn main:app --reload

Frontend:

cd frontend
npm install
npm run dev