Evidence plane
Secure ingestion, validation, storage, classification, and immutable source references.
Each layer has a defined role: receive evidence, preserve context, compute verified facts, govern investigation work, and expose only authorized results.
Secure ingestion, validation, storage, classification, and immutable source references.
Versioned deterministic calculations, detection logic, confidence, and explicit limitations.
Role-aware investigations, approvals, collaboration, reporting, and audit access.
The marketing site, the application interface, the API, the data plane, and the underlying infrastructure live as separate layers. Each is independently deployable, independently observable, and bounded by an explicit contract rather than a shared object graph. The shape of the platform is deliberate: small surfaces, clear boundaries, and observability at every layer.
That separation matters when something goes wrong. A regression in an interface layer cannot escalate into a data-corruption event, because the data layer does not trust interface inputs unconditionally; it validates against the same schema that ingestion produced. A deploy to the application layer does not change the audit format, because the audit format is owned by a different service.
Workspaces are logically isolated with strong access controls and scoped data boundaries. ReturnValue is designed to use encryption in transit and encrypted storage across core systems, with platform-managed key practices. Multiple controls sit between the public surface and anything sensitive, so failure modes are contained, and backup and recovery controls are built into the architecture.
Generative models may explain verified evidence and assist investigation, but they do not silently manufacture calculations, records, or conclusions.
Architecture should make the trustworthy path the easiest path.
Illustrative view. Values shown are examples, not customer data.