Skip to main content
Financial InstitutionsIn research · Design partners welcome

Vibrissa for financial institutions

Graph-neural-network anomaly and fraud detection built for real-time payment rails.

Instant payment systems process millions of transactions per minute, creating complex relational graphs where mule accounts, rapid fund dispersal, and synthetic identities hide within topological patterns traditional rule engines miss.

Focus Areas:

  • Dynamic topological graph representation of multi-hop account hops
  • Low-latency inference for real-time transaction screening
  • Explainable subgraph attribution for compliance and regulatory reporting
  • Zero client data sharing: models train in federated or private on-premise enclaves

This technology is currently in active research and prototyping. We do not make premature claims regarding production metrics. We are seeking forward-thinking banks and fintechs to collaborate as design partners.