Fed Aura Capital AI Workshop with Red Hat AI
Welcome to the Fed Aura Capital AI Workshop!
This workshop was created and is maintained by the Red Hat CAI Team. A big thanks to all the contributors who helped shape the content, test the labs, and provide feedback along the way - this workshop wouldn’t be possible without you.
If you have any questions or run into issues, reach out to Roberto Carratalá or Cansu Kavili-Örnek, or open an issue in the repository.
Happy learning!
What you’ll learn
In this workshop, you will build an end-to-end AI platform for mortgage lending using Red Hat AI. Through the lens of Fed Aura Capital, a fictitious mortgage lending company, you will:
Fast, Flexible, and Efficient Inferencing
-
Deploy and scale AI models - Deploy an LLM and learn how to scale it for high-throughput inferencing
-
Consume models as a managed service - Reach a shared model through the Models as a Service gateway with a scoped API key, a rate-limited subscription tier, and per-key usage metering
Connecting Data to Models and Agents
-
Shape model behavior with prompts - Use the GenAI Playground to engineer domain-specific personas and tune model responses for mortgage lending
-
Ground responses in private data with RAG - Connect your models to internal policy documents so answers reflect your organization’s rules, not generic knowledge
Accelerate Agentic AI Innovation
-
Connect agents to tools via MCP - Use MCP servers so AI agents can call real tools: affordability calculators, risk assessments, compliance checks
-
Trace and evaluate agent decisions - Use MLflow tracing to follow every tool call and LLM invocation, and run evaluations to measure quality
-
Train a machine learning model using the same platform (optional) - Use AutoML to train the loan-approval classifier model without writing model training code
Securing AI from Sandbox to Production
-
Deploy guardrails at the inference boundary - Configure guardrails to filter harmful inputs and outputs before they reach the model or the user
-
Run adversarial testing with Garak - Probe your models for prompt injection, jailbreak, and data leakage vulnerabilities before production deployment
-
Sandbox agents with OpenShell - Securely onboard AI agents with deny-by-default policies, network isolation, and audit logging
Who this is for
This workshop is designed for:
-
Platform Engineers / MLOps - who deploy and operate AI infrastructure
-
Data Scientists / AI Engineers - who build and evaluate AI models and agents
-
Application Developers - who integrate AI capabilities into business applications
-
Solution Architects - who design end-to-end AI platforms for the enterprise
Prerequisites
This workshop assumes basic familiarity with Kubernetes concepts and a general understanding of AI/ML terminology. No prior experience with Red Hat AI is required.
Workshop environment
You will work with Fed Aura Capital’s AI-powered mortgage lending platform, built with LangGraph multi-agent orchestration, Model Context Protocol (MCP) tools, developed and served on Red Hat AI.
You have several tabs available on the right hand side of the lab interface. For example, you will use OCP Console to access OpenShift and the Terminal tab to run commands.
Each exercise will provide the necessary credentials and guide you on which tabs to use.