Platform Architecture
A stronger foundation for human possibility.
Zynaut Pathways applies a shared intelligence foundation to one domain. This page describes that foundation as a discipline — the layers, the boundary around them and the capability areas — not as a running system this website contains.
Modular.Bounded.Human-Centric.
The experience presents intelligence. It is not where the intelligence lives.
- ModularSeveral distinct services, not one universal model
- Purpose-boundA declared purpose before a capability is consumed
- Provenance-carryingEvery fact keeps where it came from
- Tenant-separatedOne deployment's data never reaches another's
- ObservableA record of what was used to decide what
The big picture
How the foundation is layered.
Seven layers, read from what the system is told to what a person decides. The experience layer presents intelligence; it is not where the intelligence lives.
- Signals and evidenceWhat the system is told: declared information, verified evidence, domain events and outcomes — each carrying where it came from and under what permission.
- Governed data and knowledgeData given structure, definition and provenance, and the knowledge models that place it in context. Governance is a property of this layer, not a review after it.
- Models, reasoning and intelligence servicesDomain models, statistical and machine-learning components, rules and deterministic systems. Several distinct services, deliberately not one universal model.
- APIs and shared platform capabilitiesHow an offering consumes a capability: identity, permissions, purpose declaration and the service contract itself.
- Offerings and experience layersThe domain-specific product, platform or interface. The experience presents intelligence; it is not where the intelligence lives.
- Human and organisational decisionsA named person or accountable body acts. The system's authority ends where the brief for that use case says it ends.
- Outcomes, corrections and approved feedbackWhat actually happened, and what was wrong. Returned into the foundation only where purpose, permission and policy allow.
The governance boundary
Eight controls that surround every layer.
The boundary surrounds the architecture rather than following it. Every layer above operates inside these controls.
- PrivacyCollect what the purpose requires and no more.
- Consent and purposeA declared purpose, and a lawful basis for it.
- SecurityControls on access, transit and storage.
- Access and permissionWho may consume which capability, and on whose behalf.
- ProvenanceWhere a fact came from, and whether it is still current.
- Human oversightA person accountable for consequential outcomes.
- AuditA record of what was used to decide what.
- CorrectionA route to fix what is wrong, and to remove it.
Capability areas
Reusable intelligence services.
Named capability areas of the shared foundation. They are named here, not described as shipped features. Pathways consumes some of them; none scores, ranks or searches for a person.
- Profile intelligenceConsumed by Pathways
- Capability intelligenceConsumed by Pathways
- Opportunity intelligenceConsumed by Pathways
- Learning intelligenceConsumed by Pathways
- Institution intelligenceNot consumed by Pathways
- Financial intelligenceNot consumed by Pathways
- Workforce intelligenceNot consumed by Pathways
- Matching intelligenceNot consumed by Pathways
- Recommendation intelligenceConsumed by Pathways
- Knowledge intelligenceNot consumed by Pathways
- Decision intelligenceConsumed by Pathways
Built for what matters
Intelligence that shows its provenance.
Every capability an offering consumes is declared by purpose, checked by permission and recorded for audit. What a system is told carries where it came from; what it concludes carries what it used; what a person decides stays with the person. That is the discipline, and Pathways is being built to it.
- Purpose declared before a capability is consumed
- Provenance carried on every fact
- Distinct services, not one universal model
- Human decision as a layer of its own
- Corrections returned only where permission allows
The experience presents intelligence. It is not where the intelligence lives.
- Experience layerPathways, white-label deployments, the interfaces people use
- Intelligence servicesCapability, profile, learning, opportunity, decision — several, not one model
- Application layerIdentity, permissions, purpose declaration, service contracts
- Data and trustProvenance, consent, evidence, governance — a property of the layer
- InfrastructureCloud, security, observability
- ModularDistinct services, replaceable one at a time
- InteroperableConsumed through declared service contracts
- Purpose-boundNothing consumed without a declared purpose
- Provenance-carryingEvery fact keeps its source and its currency
- Tenant-separatedA deployment is a wall, not a folder
- ObservableA record of what was used to decide what
An architecture is a set of promises about what a system will refuse to do. Ours refuses to score a person, to reach past a permission, or to reason from a fact it cannot trace.
Questions people ask
The architecture, in plain answers
Is this a description of a running system?
No. It is the discipline Zynaut engineers to, described in layers. This website contains no backend, no data store and no model; Pathways is in development and not generally available. Nothing here should be read as a capture of software in service.
Which cloud does it run on?
Not stated, and no vendor is named. Where a deployment runs, and under which country's law its data is held, is a fact each deployment will state with its own terms rather than a claim made for the platform as a whole.
What does modular mean here?
Several distinct intelligence services rather than one universal model. Each is consumed through a declared contract, so one can be replaced, withdrawn or bounded without rebuilding the rest, and so no single component is trusted with every decision.
Are uptime, scale or performance figures published?
No. None is measured over a period a reader could check, so none is printed. When a figure is measured and verified, it will be published with its method, its dates and the conditions under which it does not hold.
How is one deployment kept apart from another?
By design at the data and access layers: a deployment's records, registries and integrations belong to it, and there is no route from one to another. An institution's view is bounded and permissioned; it is never a route to another institution's learners.
Does the architecture ever decide something about a person?
No. Human and organisational decision is a layer of its own. The system assembles evidence and reasoning; a named person or an accountable body acts, and the system's authority ends where the brief for that use case says it ends.
