Case Study

SafePay

Digital wallet MVP for Pakistan with phone/CNIC sign-up, transaction PIN, instant wallet-to-wallet transfers, and ML fraud scoring on every transfer, plus an admin panel.

See it live
SafePay
Client
SafePay
Category
SaaS · Fintech

The problem

Peer-to-peer transfers in Pakistan can be slow, and they're exposed to scams and fraud. Users usually only find out about a fraudulent transfer after the money is gone.

The solution

SafePay is a digital-wallet MVP. It scores every transfer for fraud risk in real time before the money moves. If a transfer looks suspicious (odd hour, unusual location, new recipient, amount far above the user's normal), it's blocked and the user sees why.

Key features

  • Local onboarding. Sign-up with phone (+92), CNIC, OTP verification and a 4-digit transaction PIN.
  • Wallet. Instant top-up, balance, and a full transaction history/audit trail.
  • Instant transfers. Wallet-to-wallet by recipient phone number, confirmed with the PIN.
  • AI fraud detection. An ML model scores each transaction on signals such as amount, time of day, location, device, history, recipient and behaviour. It gives a 0–100 risk score and explains why a transfer was blocked.
  • Admin panel. Real-time fraud monitoring, account freeze and decision override.
  • Security. JWT auth, bcrypt password hashing, PIN verification and ACID transactions on PostgreSQL.

Stack

React single-page app (Vite build). The footer describes the product as built with MERN + FastAPI + ML: a Node/Express API with a Python FastAPI fraud-scoring service. The homepage mentions PostgreSQL, JWT and bcrypt.