Model Governance
Transparent, auditable, clinically governed
Current Model
Version
v1.0.0
Base Model
Llama 3 8B
Fine-Tuning
LoRA
Last Updated
March 2026
5-Step Governance Process
Evidence
New data sources, benchmark improvements, or clinical gaps identified by the Clinical Advisory Committee trigger a governance cycle. Evidence is compiled including data deltas, performance metrics, and clinical relevance assessments.
Proposal
The Principal Clinical Physician (PCP) documents the current model version versus the proposed version, including data delta, benchmark comparisons, and A/B test results from shadow deployment.
Board Review
The CEO, CMO, Clinical Advisory Committee, and Technical Advisory Committee review the proposal. The CMO holds clinical veto authority over any model change that could affect patient safety.
Public Notice
Approved changes are published for a 30-day public notice period on the CH Chain Explorer and conceptualhealth.ai. All API customers receive direct notification with migration guides.
Deployment
The new model is distributed via SecureMesh. A GovernanceTransaction is written to the CH Chain containing: version, data hash, benchmark results, board signatures, and deployment date. 90-day post-deployment monitoring begins.
Emergency Rollback
In the event of a critical issue, the CEO may authorize an immediate model rollback to the previous stable version. A 24-hour retroactive public notice is issued, and the full board must ratify the rollback decision within 7 days. All rollback events are permanently recorded on the CH Chain.
Clinical Advisory Committee
The Clinical Advisory Committee provides ongoing clinical oversight of the ConceptualHealth.AI model. The committee reviews proposed model changes for clinical validity, identifies areas where the model needs improvement, and ensures that all outputs meet evidence-based medical standards. Committee members include practicing physicians, clinical researchers, and domain specialists across the 8 health axes.
Clinical Validation
Safety Oversight
Evidence Review
Quarterly Intelligence Reports
Every quarter, Conceptual Health publishes an Intelligence Report covering model performance metrics, data growth statistics, parameter changes, benchmark results, and clinical validation outcomes. Reports are available to all API customers and published on the CH Chain Explorer for public audit.
Version History
| Version | Date | Base Model | Method | Notes |
|---|---|---|---|---|
| v1.0.0 | March 2026 | Llama 3 8B | LoRA fine-tuning | Initial release. 8-axis fine-tuning on de-identified clinical data from NexusOrb platform. |