Your Infrastructure. Our Governance Engine. Zero PII Exposure.
ClarifAI's Hybrid Bring-Your-Own-Cloud architecture deploys the full governance engine inside your AWS, Azure, or GCP VPC. Production model data never leaves your perimeter. Governance intelligence does not require data surrender.
Most Governance Tools Are a Data Sovereignty Problem
Traditional SaaS AI governance platforms require you to send model predictions, feature values, and sometimes training data to their servers for analysis. For financial institutions, hospitals, and government agencies — where production data contains customer PII, sensitive financial records, or protected health information — this is not acceptable. ClarifAI was architected differently: the computation comes to your data, not the other way around.
BYOC Deployment Architecture
How the governance engine runs inside your infrastructure.
Governance Engine Deployment
The ClarifAI governance engine — including the XAI Consensus Engine, drift monitors, fairness evaluators, and audit trail — is packaged as a Helm chart and deploys to any Kubernetes cluster in your VPC. The deployment takes under 2 hours from a standard cluster.
Model Serving Integration
ClarifAI connects to your existing model serving layer — MLflow, Seldon, BentoML, SageMaker, Vertex AI, or custom REST endpoints. A lightweight sidecar or webhook captures prediction inputs and outputs without modifying your serving infrastructure.
Governance Computation (In Your VPC)
All XAI, drift, and fairness computation happens inside your infrastructure perimeter. Results — explanation scores, drift metrics, fairness reports — are stored in your data stores. Only aggregated, non-PII governance metadata is synchronised to the ClarifAI management plane for dashboarding.
Management Plane Connectivity
The ClarifAI SaaS management plane provides the governance dashboard, alert routing, and compliance reporting UI. It communicates with your BYOC engine via an encrypted, outbound-only API connection. No inbound ports need to be opened in your firewall.
BYOC Capabilities for Enterprise Deployment
Multi-Cloud Support
Deploy on AWS EKS, Azure AKS, or Google GKE. Native integrations with each cloud's IAM, KMS, and networking primitives. Multi-region and multi-cloud deployments supported.
Helm Chart Packaging
Production-grade Helm charts with configurable resource limits, horizontal pod autoscaling, and PodDisruptionBudgets. Deployment runbook provided for standard enterprise Kubernetes clusters.
MLOps Stack Integration
Pre-built connectors for MLflow, Seldon, BentoML, SageMaker, Vertex AI, and Azure ML. Custom integrations available for proprietary model serving platforms via REST webhook.
Customer-Managed Encryption Keys
All data stored by ClarifAI within your VPC is encrypted with keys you manage in your KMS. ClarifAI never has access to your encryption keys — not even for support purposes.
Air-Gap Support
For environments with no external internet access, a fully air-gapped deployment mode is available. Governance dashboards and reports are accessible within your internal network without any external connectivity.
Infrastructure Sizing Guide
Detailed sizing documentation for small (< 10 models), medium (10–50 models), and large (50+ models) deployments. Kubernetes resource manifests and cost estimates for all three major cloud providers.
Ready to Deploy Governance Inside Your Perimeter?
Schedule a 60-minute architecture review with ClarifAI's solutions engineering team.