🥗 Food Rescue Optimizer

AI-powered food rescue logistics platform — built with Java 21, Spring Boot, Timefold constraint solver, and deployed on AWS ECS.

The Problem

Over 1.3 billion tons of food is wasted globally every year while millions go hungry. The gap isn't food — it's logistics. Donors don't know where to send food. Recipients don't know what's available. And no one is optimizing the routes.

The Solution

Food Rescue Optimizer connects food donors (restaurants, supermarkets) with recipients (food banks, shelters) and uses AI to plan the most efficient pickup and delivery routes — minimizing distance, saving fuel, and ensuring food reaches people before it spoils.

How It Works

1. Donor registers and submits available food items with quantity and expiry time.
2. ML model scores spoilage risk instantly based on food category, temperature, and hours until expiry.
3. AI optimizer assigns vehicles to pickups using constraint-based optimization.
4. Routes are served via REST API for downstream apps to consume.

Key Features

• 🤖 AI Route Optimizer — Timefold constraint solver assigns vehicles to pickups across 3 hard constraints: vehicle capacity, time windows, and distance minimization.
• 🧠 ML Spoilage Scorer — Logistic regression model (Smile library) predicts spoilage risk at submission time using food category, temperature, and hours until expiry.
• ☁️ Cloud Native — Containerized with Docker, deployed on AWS ECS Fargate with RDS PostgreSQL.
• 🔄 CI/CD Pipeline — GitHub Actions automatically builds, tests, and redeploys on every push to main.
• 📋 API Documentation — Swagger/OpenAPI interactive docs with all 11 endpoints.
• 🗄️ Zero Schema Drift — Versioned Flyway migrations ensure identical database state from local to production.
• ✅ Tested — Unit tested with JUnit 5 and Mockito.

Benchmark — Solver Results

Metric	Baseline (nearest-neighbor)	Timefold Solver
Total route distance	TODO km	TODO km
Vehicles used	TODO	TODO
Solve time	—	TODO s

Sample scenario: TODO donors, TODO vehicles, TODO recipients.

Tech Stack

Layer	Technology
Language	Java 21
Framework	Spring Boot 3.4.4
AI Optimizer	Timefold Solver
ML Library	Smile (Logistic Regression)
Database	PostgreSQL + Flyway
Containerization	Docker
Cloud	AWS ECS Fargate + RDS
CI/CD	GitHub Actions
API Docs	Swagger / OpenAPI

API Endpoints

Method	Endpoint	Description
POST	/api/donors	Register a food donor
GET	/api/donors	List all donors
GET	/api/donors/{id}	Get donor by ID
POST	/api/donors/{id}/food-items	Submit a food item with ML scoring
POST	/api/recipients	Register a recipient
GET	/api/recipients	List all recipients
GET	/api/recipients/{id}	Get recipient by ID
POST	/api/vehicles	Add a vehicle
GET	/api/vehicles	List all vehicles
GET	/api/vehicles/{id}	Get vehicle by ID
POST	/api/optimize	Run the AI route optimizer

Live Demo

Resource	URL
🌐 Live API	http://3.238.51.54:8081
📖 API Docs	http://3.238.51.54:8081/swagger-ui/index.html
💻 GitHub	https://github.com/erdkash1/Food-Rescue-Optimizer

Getting Started

Prerequisites

• Java 21
• Docker + Colima (Mac) or Docker Desktop
• PostgreSQL

Run Locally

colima start
docker compose up -d
./mvnw spring-boot:run

Run Tests

./mvnw test

Demo Data

Once the seed migration is in place, a fresh clone loads the demo dataset automatically via Flyway — then call POST /api/optimize to watch the solver work.

Project Structure

src/main/java/com/foodrescue/optimizer/
├── controller/      # REST endpoints — Donor, FoodItem, Recipient, Vehicle, RouteOptimizer
├── service/         # Business logic — RouteOptimizerService, SpoilageRiskScorer (ML)
├── solver/          # Timefold constraints — FoodRescueConstraintProvider
├── domain/          # Planning entities — RoutePlan, Route, RouteStop, Vehicle, Donor, ...
├── repository/      # Spring Data JPA repositories
└── exception/       # GlobalExceptionHandler

src/main/resources/
├── application.properties.example
└── db/migration/    # Flyway — V1 init schema, V2 remaining tables, V3 spoilage fields

CI/CD

Every push to main automatically:

1. Sets up Java 21
2. Builds with mvn clean package
3. Builds the Docker image and pushes it to Amazon ECR
4. Triggers a new deployment of the ECS service

Architecture

flowchart LR
    Client -->|REST| C[Controllers<br/>Donor, FoodItem, Recipient,<br/>Vehicle, RouteOptimizer]
    C --> S[Services<br/>DonorService, FoodItemService,<br/>RecipientService, VehicleService]
    S --> ML[SpoilageRiskScorer<br/>Smile logistic regression]
    C --> Opt[RouteOptimizerService]
    Opt --> Solver[Timefold Solver<br/>FoodRescueConstraintProvider]
    Solver --> Plan[RoutePlan]
    S --> DB[(PostgreSQL<br/>Flyway migrations)]
    Opt --> DB

Loading

The whole service is packaged as a Docker image and deployed on AWS ECS Fargate, with PostgreSQL on RDS.