Security & deployment
Your security posture decides where this runs. Not us.
Andromeda deploys into the environment your security posture requires, not the one that is easiest for us to run. In the customer-deployed model, your data never traverses our infrastructure at all.
Deployment
The same context layer. Wherever your security requirements live.
Andromeda ships in two models. The difference is whose infrastructure runs it, and it determines everything downstream.
Andromeda-hosted
Runs on Amazon EKS in our AWS account. Used for R&D, demos, and pilots. Instances are pinned to us-east-1 with no cross-region replication and no data egress outside the US.
Customer-deployed
The same Terraform and Helm artifacts running inside your own AWS account, in your chosen region. Andromeda has no operational data path into these environments.
Both models
The EKS control plane and worker nodes are private. There is no public Kubernetes API endpoint, and application workloads run in private subnets behind dedicated VPC security groups.
Multi-tenant SaaS
Fastest deployment
VPC isolation, private subnets, network ACLs, security groups, least-privilege IAM, KMS encryption, and CloudTrail logging.
Dedicated Cloud
Enhanced isolation
The same controls, in an account dedicated to you.
On-Premises
Data sovereignty
TLS everywhere, managed certificate rotation, no default credentials, and no open ports beyond required services.
Air-gapped / Edge
Tactical mobility
Identical controls to cloud, plus physical security requirements documented per customer. All data stays inside your network perimeter. No telemetry, usage analytics, or crash reports leave it.
Data handling
Encryption, isolation, and residency are not settings.
Encryption
Data at rest is encrypted with AWS KMS-managed keys across RDS PostgreSQL, EBS volumes, S3, and backups, with deletion protection enforced in production. Data in transit is TLS-terminated at ingress.
Tenant isolation
Every customer holds a dedicated data encryption key. Sensitive payloads use AES-256-GCM with PBKDF2 at 100,000 iterations. Tenants are isolated at the encryption-key, identity, and database-row level. The customer-deployed model adds physical isolation at the AWS account boundary.
Residency
Andromeda-hosted instances are pinned to us-east-1 with no cross-region replication and no egress outside the US. Customer-deployed instances satisfy any residency requirement through your own region selection. Third-party data flows are limited to operational dependencies, each reviewed annually and pinned to US regions where supported.
Access control
Four layers, from the identity provider down to the document.
Identity federation
Single sign-on via SAML 2.0 and OIDC with any enterprise identity provider, including Okta, Azure AD, Keycloak, and Auth0.
Role-based access
Platform administrators, tenant administrators, editors, and viewers, with configurable permission boundaries.
Document-level permissions
Attribute-based access control allows fine-grained restrictions on individual documents or document sets.
Audit trail
Every data access event is logged with user identity, timestamp, resource accessed, and action performed.
Enforcement
Every API route enforces authentication, role and ACL checks, and rate limiting. All infrastructure is defined in Terraform, with no manual changes in production.
Frameworks and audit readiness
We say what we hold and what we do not.
The platform is designed against SOC 2, NIST 800-53, HIPAA, and FIPS 140-2 control frameworks.
Andromeda does not currently hold formal certification against these frameworks. Broader audit-readiness is on the 2026 roadmap. Where a control is implemented, it is documented. Where certification is pending, we say so.
What is in place today
- Automated compliance scanning in CI/CD pipelines
- Immutable audit logs with cryptographic integrity verification
- Quarterly internal security reviews and annual third-party penetration testing
- Documented incident response procedures with defined SLAs by severity
- Change management with approval workflows and rollback procedures
Bring us your posture requirements.
Deployment mode is usually settled before capability is. Tell us the constraint and we will tell you whether the platform meets it.