Microsoft Azure User Group - Chennai

Large Language Models (LLMs) have revolutionized AI, enabling systems that can reason, plan, and generate human-like responses. Yet, when organizations move from prototypes to production, they quickly discover that a powerful model alone is not enough.

In this session, we'll use Microsoft Azure AI Foundry to build and orchestrate agents powered by both GPT and Claude models, demonstrating how modern organizations can leverage multiple foundation models within a unified enterprise AI platform.

How does an AI agent securely access enterprise data?
How does it remember previous interactions?
How does it decide when to call a tool, seek human approval, or recover from failures?
How do we monitor, govern, and trust autonomous systems operating in real-world environments?

The answer lies in the Agent Harness—the orchestration layer that surrounds the model and enables reliable execution.

In this session, we'll explore how modern AI agents are built using memory, context management, tool orchestration, governance, observability, and safety controls.

Through practical architecture patterns and live demonstrations, you'll learn how to transform AI models into production-ready autonomous systems capable of delivering real business value.

Whether you're a developer, architect, engineering leader, or AI enthusiast, this session will provide practical insights into building secure, scalable, and trustworthy AI agents.

## What You'll Learn

### Understanding Agent Harness Architecture

• What is an Agent Harness?
• Why models alone are insufficient
• The evolution from chatbots to autonomous agents

### Core Building Blocks

• Context Engineering
• Memory Management
• Tool Calling & Function Execution
• Planning & Reasoning Loops
• Human-in-the-Loop Systems

### Reliability & Governance

• Guardrails and Safety Controls
• Approval Workflows
• Error Recovery Strategies
• Observability and Monitoring
• Cost and Token Management

### Enterprise Integration

• Connecting Agents to SQL Databases
• Knowledge Retrieval with RAG
• API and Service Integration
• Business Process Automation

### Emerging Trends

• Model Context Protocol (MCP)
• Multi-Agent Systems
• Agent-to-Agent Communication
• Enterprise Agent Governance

## Who Should Attend?

• Software Developers
• AI Engineers
• Cloud Engineers
• Solution Architects
• Technical Leads
• Engineering Managers
• CTOs and Technology Leaders
• Anyone interested in Agentic AI

***