AI quickstart: Protecting inference with F5 Distributed Cloud and Red Hat AI

Earlier this year, we launched the Red Hat AI quickstart catalog, a collection of ready-to-run blueprints designed to help organizations move from talking about AI to using large language models (LLMs) to solve real-world problems. This provides systems integrators and architects with example AI solutions that Red Hat engineering has tested and streamlined for easy deployment.

Once you've successfully rolled out an interactive solution on Red Hat AI, however, the next question is usually, "How do I protect this in the real world?"

To help answer this, we've expanded the AI quickstarts catalog with one of our first partner-led entries: The F5 Distributed Cloud API Security AI quickstart.

Protecting your AI endpoints

Most organizations have no trouble spinning up a basic chat assistant or a retrieval-augmented generation (RAG) demo. The friction starts when they realize that an inference endpoint is, at its core, an API. APIs are the primary target for modern exploits.

For those of us helping customers architect these systems, security concerns are often what prevent promising pilots from reaching production. This new AI quickstart, collaboratively developed by F5 and Red Hat, helps you get past that hurdle. It demonstrates how to apply enterprise-grade protection before users begin interacting with your AI models.

Inside the F5 Distributed Cloud API Security AI quickstart

The F5 Distributed Cloud API Security AI quickstart is a modular blueprint that integrates F5 Distributed Cloud (XC) Services with the Red Hat AI platform. It's designed to be deployed in under 90 minutes, giving you a fully functional, protected environment to demonstrate:

• Schema validation: So your LlamaStack or vLLM endpoints only process well-formed, authorized requests
• Sensitive data guardrails: Automatically detecting and redacting personally identifiable information (PII) or proprietary data before it ever leaves your environment
• Resource protection: Implementing rate limiting and bot defense so your GPU cycles are used by legitimate users, not malicious scrapers
• Hybrid flexibility: Whether your model is running on-premises or in a public cloud, the architecture remains consistent

Building together

By bringing F5's decades of security expertise to an AI quickstart, we're demonstrating a reusable method  for addressing many of these "Day 2" problems.

The goal isn't just to kick the tires, it's to provide a predictable, reusable framework so that when a customer asks how their data will be protected, you'll have a working, demonstrable response.

Get started

You can clone the repository from GitHub and take it for a test drive on your cluster today: Explore the F5 API Security quickstart.