Deploy a Production-Ready NVIDIA AI-Q Blueprint on Oracle Cloud Infrastructure

NVIDIA and Oracle Detail Production AI Agent Deployment

NVIDIA and Oracle have published an open-source reference architecture for deploying the NVIDIA AI-Q multi-agent system on Oracle Cloud Infrastructure (OCI). This guide gives developers and platform engineers a direct, production-ready path to run sophisticated, long-horizon AI agents on enterprise-grade cloud infrastructure, moving beyond the limitations of local development environments.

Technical Architecture on OCI

The deployment methodology separates infrastructure from the application using standard DevOps tools. Infrastructure components on OCI are provisioned using Terraform, while the application workloads are installed on the Oracle Kubernetes Engine (OKE) via a Helm chart from NVIDIA's NGC registry. The architecture is designed for repeatability and clean teardown with a single command.

• Infrastructure (Terraform): Creates the Virtual Cloud Network (VCN), OKE cluster, public Load Balancer, and OCI Vault for storing API keys securely.
• Application (Helm): Deploys three main containerized workloads on OKE: a FastAPI-based backend for the agent logic, a Next.js frontend UI, and a PostgreSQL database for state management.
• Prerequisites: Requires an OCI tenancy with sufficient service limits, API keys for NVIDIA NGC and Tavily, and local tools including Terraform, kubectl, and Helm.

From Local Prototype to Cloud Production

This blueprint addresses a critical operational gap by providing a standardized, infrastructure-as-code solution for complex AI agent deployment. By codifying the setup on OCI, the collaboration between NVIDIA and Oracle lowers the barrier for organizations to move from experimental agentic AI concepts to scalable, managed, and secure production systems. The use of extensible tools like the NeMo Agent Toolkit further provides a pathway for enterprises to customize and adapt these agents for specialized business functions.

This collaboration provides a prescriptive, infrastructure-as-code blueprint that moves complex AI agent deployment from experimental local setups to a repeatable, enterprise-grade process on public cloud, directly addressing a key operational hurdle for production AI.