ORIGIN: The Evidence Layer for Regulated AI
The platform infrastructure behind AI-originated clinical endpoints. Veterinary pruritus is the first deployed endpoint model — with audit trails, reproducibility evidence, and complete model lineage built in.
AI Compute Access Fund — Government of Canada
Selected as one of 44 Canadian companies in the first cohort of the AI Compute Access Fund — launched under Canada’s Sovereign AI Compute Strategy to help SMEs access the high-performance compute needed to build and scale AI.
OneKind AI was selected as a recipient of the AI Compute Access Fund administered by Innovation, Science and Economic Development Canada (ISED). The ORIGIN platform is the infrastructure that qualified — a sovereign AI trust stack for regulated model governance with Canadian data residency as the default.
The AI Continuity Problem — and How the ORIGIN Platform Solves It
The gap between what AI systems produce and what can be proven, explained, and reproduced.
Manual Documentation Fails
Model versions tracked in spreadsheets. When people leave, critical context disappears.
Models Replaced, Not Archived
Old versions deleted or lost. Auditors asking "what model was used?" get silence.
No Data-to-Decision Lineage
Data, pipelines, and outputs in separate silos. Reconstructing a decision path takes weeks.
Permanent Archive
Every data version, model artifact, and processing step stored permanently. Nothing overwritten.
Complete Versioning
Models, datasets, and configurations versioned automatically. Roll back to any point.
Built-In Auditability
Every action logged with full context. Regulatory-ready audit trails generated automatically.
Platform Blueprint
Six integrated layers. Each independently versioned, auditable, and replayable.
Veterinary Pruritus Endpoint Model
The first clinical endpoint model deployed on the ORIGIN platform. Captures scratching, biting, rubbing, and head shaking as continuous, structured behavioral events for companion animal dermatology trials.
See study integration details →Inside Each Layer
What each layer does, and how the ORIGIN platform answers the questions regulators ask most.
Video Integrity Platform
Data provenance, timestamp integrity, and secure transfer from the point of capture. Every frame is cryptographically traceable before inference begins.
How do we know this video was captured correctly?
The ORIGIN platform validates capture parameters, timestamps, and file integrity at ingestion. Every frame is checksummed and linked to its source metadata before processing begins.
Behavioral Analysis Engine
Frame-by-frame detection and classification of pruritic behavioral events. Deterministic inference — same video, same model version, same result, every time.
Can we reproduce Trial 482's results from 18 months ago?
Yes. The ORIGIN platform preserves the exact environment snapshot, including library versions, Docker images, and configuration files.
Regulated Model Operations
Controlled model releases. No silent retraining. No automatic threshold shifts. Every model change is versioned and logged.
Why did the model's output change between v2.3 and v2.4?
The ORIGIN platform shows the exact dataset additions, training parameter changes, and code commits between versions.
Adverse Events + Human Review
AI-originated behavioral signals structured for human confirmation before entering pharmacovigilance workflows. Formalizing where AI ends and human judgment begins.
Which behavioral signals triggered this flag, and who confirmed it?
The ORIGIN platform traces every flagged event back to the originating model output, and logs the reviewer identity, timestamp, and disposition for each confirmation.
Regulatory Evidence Architecture
Endpoint definition versioning, validation artifact vault, threshold justification log, and reproducibility inspection replay.
Can we reconstruct the full evidence chain for this endpoint?
Yes. The ORIGIN platform links every endpoint value back through its threshold definition, model version, input data, and validation artifacts — a complete chain of custody from raw data to final result.
API integration layer → EDC systems (Medrio, Prelude, Vault EDC, Medidata Rave) + pharmacovigilance (Argus, Vault Safety)
Security & Data Sovereignty
Configurable security layers — from cloud-hosted to air-gapped on-premises deployments.
Base Security
Enterprise encryption, RBAC, and SOC 2-aligned infrastructure on Google Cloud. Included in every deployment.
Data Residency Controls
Region-specific storage (EU, US, Canada) for data sovereignty. GDPR and CCPA compliant. Configurable jurisdictions.
Trusted Execution Environments
Hardware-protected enclaves. Data encrypted even during processing. Zero-access computing.
NVIDIA FLARE Integration
Federated learning across multiple sites without sharing raw data. Multi-site collaboration with complete data sovereignty.
Built for Every Party in the Submission Chain
For Regulatory Reviewers
Every AI-generated endpoint ships with run ID, model version, dataset hash, and reproducibility evidence — so reviewers can verify results without manual reconstruction.
Regulatory Documentation →For CRO Quality Teams
Sovereign deployment, SOC 2 aligned security, and structured EDC-ready outputs — so CROs can offer AI-enhanced annotation without taking on regulatory risk.
CRO Partnership Details →Ready to Make Your AI Defensible?
The ORIGIN platform provides the documentation infrastructure that makes regulated AI trustworthy.