Infrastructure

Private Execution Fabric

Deploy MazeLabs within customer-controlled environments. From managed SaaS to private VPC to fully local execution — your data never leaves your boundary.

Deployment Models

Cloud SaaS

Available

Fully managed MazeLabs deployment. Customer data is encrypted at rest and in transit. Strict tenant isolation with no cross-tenant data access.

  • SOC 2 compliance path
  • Tenant-isolated storage
  • 99.9% uptime SLA
  • Automatic updates

Private VPC

Available

MazeLabs deployed within the customer's AWS, GCP, or Azure VPC. No data leaves the customer's network boundary. Customer manages encryption keys.

  • Customer-managed KMS
  • VPC-native deployment
  • Private endpoints only
  • Customer-controlled networking

Local Execution

Available

Run MazeLabs entirely on-premises. Docker-based deployment with local model execution. Zero external network calls for simulation and scoring.

  • Docker / K8s deployment
  • Local model execution
  • Zero external calls
  • Full data sovereignty

Air-Gapped

Planned

Fully disconnected deployment for regulated environments. Complete offline operation with manual update packages and local model inference.

  • No internet required
  • Manual update packages
  • Offline model inference
  • ITAR / FedRAMP path

Deployment Comparison

CapabilityCloud SaaSPrivate VPCLocalAir-Gapped
Data leaves customer networkYes (encrypted)NoNoNo
Customer-managed encryptionOptionalYesYesYes
External AI model callsConfigurableConfigurableLocal onlyNone
Automatic updatesYesYesManualManual
Audit logging
Redaction engine
Offline operationNoNoYesYes
Compliance pathSOC 2SOC 2, HIPAASOC 2, HIPAAITAR, FedRAMP

Privacy Controls

Customer-Owned Data

All operational evidence, simulation state, and scoring data is owned and controlled exclusively by the customer.

Encryption at Rest & Transit

AES-256 encryption for stored data. TLS 1.3 for all data in transit. Customer-managed encryption keys in VPC and local deployments.

Audit Trails

Every data access, model invocation, simulation event, and administrative action is logged with immutable audit trails.

Model Routing Boundaries

Customers define which AI models can be used, which data can be sent to external APIs, and which operations must remain local.

Redaction-First Pipeline

All evidence passes through the Redaction Engine before any AI reasoning. Secrets, PII, and internal infrastructure details are masked by default.