Product · Private LLM Deployment
Roe runs dedicated LLM weights inside your own cloud account. Inference happens on your GPUs, behind your firewall. PII never leaves your perimeter. No OpenAI, no Anthropic, no shared model plane.
What you get
Every private llm deployment deployment ships with the same evidence-first reasoning, audit trail, and policy fit as the rest of Roe.
Your tenant gets its own model instance. Weights, GPUs, and inference cache are isolated. No batching across customers. No shared model plane.
Deploy in AWS, GCP, or Azure inside your own cloud account. Networking, IAM, and KMS scoped to your security baseline. No outbound API calls to vendor infra.
Sensitive fields like SSN, PAN, and DOB are tokenized at ingestion. The model reasons over redacted tokens. Plain-text never crosses your network boundary.
KMS, Cloud HSM, or Azure Key Vault. You hold the keys for data at rest, in transit, and in the inference cache.
How it works
A Terraform module spins up Roe and dedicated LLM weights inside your cloud account. IAM, networking, and KMS map to your security baseline.
Roe reads from your warehouse and risk providers over private endpoints. Nothing routes through the public internet.
Inference runs on dedicated GPUs in your tenant. Audit logs land in your SIEM. Roe is just another workload your security team controls.
Bytes leaving your VPC
LLM inference plane
Customer-managed encryption keys
From provisioning to first case