Risk decisioning infrastructure with governed rules, recorded decisions and verifiable evidence.
QUORUM converts real-time risk signals into governed decisions using per-institution rules and behavioural scoring, with AI analysis kept off the real-time path. Every decision is recorded before it is returned and can be proven to a third party.
System Architecture
System Function
QUORUM processes continuous event streams and resolves them into controlled outcomes under governance constraints.
AI output never issues a verdict on its own.
- Ingests identity, transaction, device, and behavioral signals
- Maps each request to rule facts, leaving out anything it does not carry
- Applies governed rule arbitration across policy hierarchies
- Combines rule results with a statistical behavioural score
- Resolves deterministic outcomes under execution constraints
- Records each decision before release and seals it into a signed ledger
Decision Model
Each recorded decision can be verified independently.
Design Constraints
Governance precedence
Institution-defined rules decide; AI model output never overrides them.
Deterministic execution
The real-time path is a rule engine: the same inputs under the same rule set produce the same verdict. AI analysis runs separately and is never on that path.
Explainability requirement
Every decision returns the rules and signals behind it, and every block gets a signed counterfactual certificate.
Tamper-evident audit
Decision records are hash-chained and signed, so any later alteration is detectable without trusting QUORUM. Records hold a digest of the inputs, not the raw inputs.
Capabilities
Decision Orchestration
REST, ISO 8583, message-queue and flat-file channels feeding one rule engine, failing closed when evaluation is impossible.
Policy Control
Per-institution rule sets, shadow mode for new rules and operator promotion with optional FROST approval.
Behavioral Intelligence
Typing, pointer, touch and motion signals, automation indicators and per-user baselines, scoped to each institution.
Risk Modeling
Rule-based risk scores with the fired rules and signals attached, plus a signed counterfactual certificate for every block.
Human-in-the-Loop
Analyst review of AI findings, WebAuthn step-up for privileged actions and multi-party (FROST) approvals for rule changes.
Forensic Continuity
Tamper-evident decision logs, Merkle inclusion proofs and zero-knowledge score proofs for every recorded decision.
Deployment Architecture
Container images deployed on-premise or in the institution's private cloud. AI models run locally; no customer data leaves the institution.
Deployment Patterns
QUORUM runs inside the institution's own environment, on-premise or in its private cloud, with the same decision and audit behaviour in every deployment.
Examine Deployment Architecture
Observe Governing Logic
In Real-Time
Execute stress-test scenarios, evaluate policy-driven arbitration, and and watch a decision object build step by step. The simulator is illustrative and runs in your browser.
Illustrative simulation // Runs in your browser