VeriPay

Short overview

VeriPay is a web application that helps organizations upload invoices (PDF or image files), extract key information using AI-assisted document understanding, and verify whether an invoice is trustworthy using anomaly detection, vendor validation, and document integrity checks.

It combines rule-based validation, LayoutLMv3-based document understanding, Isolation Forest anomaly detection, and optional LLM-assisted extraction via Ollama.

---

Problem it solves

Manual invoice verification is:

• Slow
• Error-prone
• Vulnerable to fraud

VeriPay provides:

• Automated invoice analysis
• Vendor and bank verification
• Audit-ready tracking
• A structured review workflow

---

Who this project is for

Audience	Why they use VeriPay
Developers / students	Learn full-stack + AI integration
Finance teams	Review invoices with risk signals
Security-focused teams	Detect fraud and anomalies

---

Key features

• 🔐 User authentication (session-based)
• 📄 Invoice upload (PDF, PNG, JPG)
• 🧠 AI anomaly detection (Isolation Forest + LayoutLM)
• 🏦 Vendor + bank verification
• 📊 Dashboard with stats and recent invoices
• 📝 Human review system
• 📁 Audit logs
• 🤖 Optional LLM extraction via Ollama

---

Tech stack

Technology	Role
Next.js / React	Frontend
TypeScript	Frontend typing
Tailwind CSS	UI styling
FastAPI	Backend API
SQLAlchemy	ORM
PostgreSQL	Database
Alembic	Migrations
PyTorch / Transformers	AI models
scikit-learn	Anomaly detection
Ollama	Local LLM
Docker Compose	Full system setup

---

Project structure

VeriPay/
├── backend/         # FastAPI backend
├── frontend/        # Next.js frontend
├── ai_pipeline/     # ML models
├── docker-compose.yml
└── README.md

---

Installation guide (Docker)

1. Install Docker

Install Docker Desktop and ensure it is running.

---

2. Clone the repository

git clone https://github.com/VeriPay-Project/VeriPay.git
cd VeriPay

---

3. Run the project

docker compose up --build

This starts:

• Frontend → http://localhost:3000
• Backend → http://localhost:8000
• Ollama → http://localhost:11434

---

4. Open the app

http://localhost:3000

---

5. Stop the project

docker compose down

---

How it works

1. User uploads invoice
2. Backend extracts text + layout
3. AI models analyze structure and anomalies
4. Vendor and bank validation runs
5. Results are stored and shown in dashboard

---

API

Base URL:

http://localhost:8000

Interactive docs:

http://localhost:8000/docs

---

Example workflow

1. Register or login
2. Create vendor
3. Upload invoice
4. Run analysis
5. Review results
6. Approve or escalate

---

Database

• PostgreSQL stores:

  • Users
  • Invoices
  • Vendors
  • Analysis results
  • Audit logs

---

Common issues

Issue	Fix
Docker not running	Start Docker Desktop
Slow first run	Models are loading (normal)
API not responding	Check backend logs
No data showing	Ensure user is logged in

---

Future improvements

• Advanced fraud scoring
• Real-time analytics
• Multi-tenant support
• Cloud deployment

---

License

MIT License

---

Contact

Maintained as part of a university capstone project.