The AI infrastructure
CRATON.
Built to stay. Built to stay yours.
CRATON, the Criticalinsights™ sovereign stack: hardware and models inside the institution, connected by a governed federation layer.
Named for the craton: the ancient core of a continent, stable for billions of years, the ground everything else is built on. Canada sits on one of the largest on Earth.
Sovereign by design, enterprise by default
Bring the compute to the data, not the data to the compute.
Most capable AI runs in third-party cloud services where patient information cannot go, and most hospital systems were never designed for modern AI workloads. CRATON closes that gap with hardware and models inside the institution, connected by a governed federation layer.
This is the architecture. SALVEA is the governed platform that runs on it, the control layer for every assistant, agent, and workflow the hospital deploys.
Model & product agnostic
Deploy any AI. Build your own. All inside your guardrails.
CRATON and SALVEA are model- and vendor-agnostic. Run any AI solution from any provider, or develop and deploy your own models and products, every one of them governed, audited, and held to the policies the institution sets. Your guardrails, not ours.
Engineered for the standards North American healthcare runs on
- HL7 FHIR
- SNOMED CT
- LOINC
- ICD-10-CA / ICD-10-CM
- National drug identifiers (DIN / NDC)
Provincial AI hub
A dedicated sovereign hub serving every hospital in the province. Institutions de-identify locally and send data up for province-governed model development and certified access. Interconnect-ready, so provincial hubs can federate nationally if required.
Institutional infrastructure
Compute inside hospital walls: existing technology staff plus one or two roles, no per-user fees, no cloud dependency. Deployed on a secure network segment with no inbound internet exposure.
Unit and department nodes
Local nodes running small and large language models we deploy and govern, sized to a clinical unit or department. Compact workstations and rack nodes rather than construction projects.
Edge AI at the point of care
Small models connected to monitors and record systems at the bedside for sub-second analysis, with device-data gateways feeding the unit tier.
Sovereign by design, enterprise by default
Governance
Data residency follows stewardship, and the ontology federates.
Local by default
Patient information stays with the organization legally responsible for it. Every AI response for clinical work is generated inside the institution.
Provincial when authorized
De-identified, governed contributions support model development and research under provincial authority, never as an automatic data transfer.
National when strategically required
The architecture is interconnect-ready, so provincial capacity can join a national fabric without redesigning what is already running.
Design commitments
01 / No third-party control
No third-party operator, custodian, key-holder, or licensor controls the system. Hardware is installed and operated on premises, under the institution's control, in Canada or the United States.
02 / Governed models
Small and large language models we deploy and govern on premises, so capability does not depend on an external vendor's terms.
03 / Existing operations
Hospital network segmentation and operation by the teams already running institutional systems.
04 / Staged adoption
Deployment moves from lowest to highest sensitivity: knowledge assistance first, documentation support next, then operational forecasting and monitoring intelligence.
Deployment models
The top tier follows who governs the hospital.
In Canada the governing body is a public health authority under provincial law, so the fourth tier is a province-governed hub. In the United States it is the health system that owns the hospitals, often across several states, so the fourth tier is an enterprise hub the system governs itself. The lower three tiers are identical.
One sovereign foundation. Everything governed.
Criticalinsights builds the foundation. Clinical AI runs on top of it: ours, the hospital's own, or any vendor's, all governed the same way.