PayPal is going all-in on AI — and PayRoute is the AI decision layer for payments. Every transaction is intelligently routed across PayPal + 4 channels, guarded by explainable, real-time fraud defense.
PROBLEM: Small merchants struggle integrating multiple payment channels — single-channel failures drop orders, fees vary wildly, and there's no unified fraud layer.
SOLUTION (PayPal + AI):
- PayPal deep integration: POST /api/v1/paypal/order connects directly to PayPal Sandbox Orders API v2 (OAuth2 client-credentials flow, returns real approval URL). Without credentials it gracefully falls back to mock mode so judges can run it instantly.
- Explainable AI fraud detection (payroute-fraud-v2): feature engineering + weighted logistic regression + adaptive z-score anomaly detection. Every response includes per-feature contribution weights — fully interpretable. Scores 0-100, auto-blocks at 80+.
- Smart routing: channels scored by success_rate*100 − fee_cost − latency_penalty, with automatic failover to backup channels.
TECH: Pure Golang (stdlib net/http, zero heavy deps), Docker one-liner, native HTML dashboard.
WHAT'S NEXT: Learn routing weights from live traffic, add a persistent transaction ledger, and connect real Stripe/Adyen channels — evolving the AI decision layer from demo to production.
Try it in 3 minutes: go run ./cmd/server → http://localhost:8080. To see the AI in action, POST /api/v1/pay with amount 9999 — watch it get blocked with per-factor weights.
Built With
- artificial-intelligence
- fintech
- go
- machine-learning
- paypal
- rest-api
Log in or sign up for Devpost to join the conversation.