I already run all payments through an in-house service we simply call “paymentgateway.” Before funds leave an account I now want an AI decision layer that can instantly approve, flag, or decline a request based on two new rule sets: • Transaction limits – the model needs to recognise and enforce daily, per-transaction, and monthly caps that can vary by user tier and currency. • User eligibility – it should check a customer’s current status (KYC passed, account age, risk score, etc.) and decide whether the transaction can proceed. The gateway is live, so the AI must slot in without disrupting existing calls to our API. Right now we only interface with our own service, but future connectors to PayPal, Stripe, or Square should stay possible, so keep the design provider-agnostic. My preferred flow is: payment request → AI rules engine → gateway authorisation. I will expose webhooks or a middleware endpoint for you to hook into. You are free to work in Python, Node, or another performant language as long as the solution comes back as a RESTful micro-service (Docker container is ideal). Deliverables 1. Rule-based / ML model that evaluates the two conditions in <200 ms 2. API endpoints (JSON over HTTPS) with authentication 3. Integration guide and sample code that shows the call sequence with success, hard fail, and soft-fail responses 4. Unit and load tests demonstrating correct handling of the daily, per-transaction, and monthly limits plus user eligibility scenarios Acceptance criteria: end-to-end test payment hits the AI layer, returns the appropriate decision, and logs the outcome to our dashboard. If you’ve done similar compliance or risk engines, let me know—especially anything tied into multiple gateways.