Question 1

What is GRYPI?

Accepted Answer

GRYPI is a real-time payment risk orchestration and fraud intelligence platform for PSPs, acquirers and gateways. It combines 150+ fraud vendors, a 5-model machine-learning ensemble, AML/sanctions screening and device fingerprinting into a single API call that returns a risk decision in under 300 milliseconds.

Question 2

How does GRYPI detect payment fraud?

Accepted Answer

GRYPI extracts 107 enrichment signals from each transaction with zero added latency, runs a parallel fan-out across 150+ fraud-intelligence vendors, scores the result with a 5-model ML ensemble, and checks a cross-PSP entity network so fraud confirmed at one provider is flagged before it reaches another. It returns an ALLOW, REVIEW or BLOCK decision with named reason codes.

Question 3

Does GRYPI support server-to-server and hosted checkouts?

Accepted Answer

Yes. GRYPI infers the checkout channel from each request — hosted payment page, inline collection, or pure server-to-server — and scores it correctly with no per-merchant configuration. One endpoint covers every integration a PSP's merchants run.

Question 4

What is the GRYPI device fingerprinting SDK?

Accepted Answer

It is a keyless one-line JavaScript SDK embedded on the checkout page that captures device, browser and behavioural signals. Bot and confidence verdicts are computed server-side — nothing on the page can be reverse-engineered or spoofed — and the signals are folded into the risk score before any external vendor is called.

Question 5

How fast is GRYPI?

Accepted Answer

GRYPI returns risk decisions at roughly 287ms p99 latency, with 107 enrichment signals extracted in-payload, per-provider circuit breakers and a 2.5-second hard timeout.

Question 6

Is GRYPI compliant with data-protection regulations?

Accepted Answer

GRYPI's cross-PSP entity network uses k-anonymous SHA-256 entity broadcasts so raw personal data never leaves a tenant, and device fingerprinting is value-blind and PCI-safe. The platform is designed to be GDPR-clean.