Zymr deployed an AI-based fraud detection engine using ML models trained on transaction patterns, behavioral profiles, and device signals. The system was built on Google Cloud using TensorFlow and BigQuery. Alerts were streamed via Pub/Sub and visualized using Looker dashboards. Automated workflows triggered AML case management, while STR submissions were logged and version-controlled for audit readiness. We implemented granular RBAC and encrypted audit trails for regulatory compliance.
A fast-scaling digital wallet and payment solutions provider with 15M+ users across Southeast Asia. Facing increasing fraud attempts and expanding regulatory requirements, the client sought to build a real-time fraud detection system with robust reporting and governance controls.
The existing fraud prevention system was rules-based and reactive, with high false positives and delayed detection. Transaction data was siloed, lacking enrichment and central visibility. Regulatory authorities demanded timely Suspicious Transaction Reports (STRs) and audit logs, which were maintained manually.
Fraud detection accuracy improved by 65%, and report submission time was cut by over 90%. Regulatory compliance became continuous and traceable.
Zymr deployed an AI-based fraud detection engine using ML models trained on transaction patterns, behavioral profiles, and device signals. The system was built on Google Cloud using TensorFlow and BigQuery. Alerts were streamed via Pub/Sub and visualized using Looker dashboards. Automated workflows triggered AML case management, while STR submissions were logged and version-controlled for audit readiness. We implemented granular RBAC and encrypted audit trails for regulatory compliance.
Show More