Governance & Security AI
HIPAA-compliant deployment with full audit and governance. SOC 2 Type II, HIPAA BAA, 99.9% uptime, complete model lifecycle control.
Product capabilities
Citadel wraps all Mergic products with private deployment, policy controls, audit trails, and model versioning — so you can run AI on PHI with full compliance and IP protection.
Deploy NVIDIA NIM inference microservices in your VPC, on-prem, or air-gapped environment — all Mergic products run on private endpoints with full model control. PHI never leaves your infrastructure, and you retain ownership of all fine-tuned weights and inference logs.
Enforce role-based access, data residency policies, and PHI masking at the inference layer — all logged and auditable.
Every model prediction, retrieval, and clinical decision is logged with full provenance for compliance reporting.
Track model versions, weights, training data lineage, and performance metrics across all Mergic products.
Fine-tune foundation models on your proprietary data with NVIDIA DGX — models stay in your boundary.
Deploy Mergic products in fully air-gapped environments with no external connectivity for maximum security.
NVIDIA healthcare stack
Self-hosted inference microservices for all Mergic products — deploy in VPC, on-prem, or air-gapped with full model control.
Fine-tune foundation models on proprietary health data with DGX infrastructure — models and data stay in your boundary.
Performance metrics
Full compliance with SOC 2 Type II controls for security, availability, and confidentiality
Business Associate Agreement with complete PHI safeguards and breach notification protocols
Uptime SLA for production AI deployments with 24/7 monitoring and incident response
Audit coverage — every prediction, retrieval, and clinical decision logged with full provenance
PHI leaves your infrastructure — all inference, training, and storage in your security boundary
Use cases
Deploy all Mergic products (Vision, Helix, Lumen, Forge, Atlas) behind Citadel governance in your VPC or on-prem. All inference runs on private NIM endpoints, PHI is masked at the policy layer, and every prediction is audited for compliance reporting. Fine-tune models on your proprietary data with DGX infrastructure — models and weights stay in your boundary. Health systems get production AI with SOC 2 Type II and HIPAA BAA coverage.
Deploy Mergic in fully air-gapped environments with no external connectivity for maximum security and IP protection.
Run AI on sensitive cohorts with policy-enforced data residency — PHI stays in-region with full audit trails.
Deploy Vision, Helix, and Lumen in clinical production with full model governance — track versions, weights, training data, and performance metrics. Roll back to previous versions if performance degrades or new safety concerns emerge.
Run Forge on proprietary targets and compound libraries with full IP protection — models and data stay in your VPC.
How it works
Citadel wraps all Mergic products with private inference, policy enforcement, audit logging, and model governance for production AI in healthcare.
Install Citadel control plane in your VPC or on-prem; provision private NIM endpoints for all Mergic products
Set role-based access, data residency rules, PHI masking policies, and fine-tuning permissions
All Vision, Helix, Lumen, Forge, Atlas inference routed through Citadel policy and audit layer
Log every prediction, retrieval, and clinical decision with full provenance; track model versions and weights
Real-time monitoring for model drift, policy violations, and performance degradation with alerting
Generate audit reports for SOC 2, HIPAA, and internal compliance reviews with full prediction lineage
Integration
Integrate with Okta, Azure AD, Google Workspace, or custom SAML/OIDC for role-based access
Export audit logs to Splunk, Datadog, ELK, or custom SIEM for centralized security monitoring
Deploy on AWS, Azure, GCP, or on-prem with Kubernetes, Terraform, or custom infrastructure-as-code
Related products
Real-time GPU radiology and pathology AI for detection and segmentation.
Whole genomes from a day to under an hour with Parabricks-backed secondary analysis.
The clinical-decision agent that sees across modalities — imaging, genomics, EHR.
Generative AI for protein structure prediction and molecular design with BioNeMo.
The consented, de-identified data layer that powers the Mergic flywheel.
FAQ
Answers for CISOs, compliance officers, and IT leaders evaluating Citadel for production AI.
Yes. Citadel deploys as a control plane in your VPC, on-prem data center, or air-gapped environment. All NIM inference endpoints run in your infrastructure, and PHI never leaves your security boundary.
Citadel enforces PHI masking at the policy layer, logs every prediction with full provenance, and supports Business Associate Agreements (BAA). All inference, training, and storage stays in your HIPAA-compliant infrastructure.
Yes. Citadel provisions NVIDIA DGX infrastructure in your boundary for fine-tuning foundation models on your proprietary health data. Models, weights, and training data never leave your security perimeter.
Every model prediction, data retrieval, and clinical decision is logged with full provenance — input data, model version, output, timestamp, user. Audit logs export to SIEM or custom compliance systems for SOC 2 and HIPAA reporting.
Get started
Bring your security policies, data residency rules, and audit needs. We'll map the deployment architecture, governance controls, and compliance coverage.