Agentic AI

Agentic AI engineers design, build, and operate AI agents that plan, use tools, remember context, and finish real business tasks.

Agentic AI engineers design, implement, and operate AI agents. They build planning loops, tool calling, memory, multi-agent workflows, guardrails, and secure Azure deployments so agents can finish end-to-end tasks safely.

Our course covers these roles extensively

• Customer Support Agents: Agents that read tickets, call CRM tools, and draft safe replies.
• Research and Knowledge Agents: Agents that search docs with RAG, summarize, and cite sources.
• Ops and DevOps Agents: Agents that check logs, open tickets, and run approved runbooks.
• Sales and Ops Assistants: Agents that qualify leads, update CRM, and prepare briefs.
• Multi-Agent Teams: Planner, researcher, and executor agents working together.
• Secure Enterprise Agents: Agents with identity, permissions, audit logs, and guardrails.

Orchestration Frameworks

1. LangGraph:

• Build stateful agent graphs with planning, branching, and human-in-the-loop steps.
• Great for complex workflows and multi-step tools.

1. Microsoft Agent Framework:

• Microsoft tooling for building and running enterprise agents.
• Fits well with Azure identity and enterprise controls.

1. Microsoft Foundry Agent Service:

• Managed agent service in Microsoft Foundry for hosting and operating agents.
• Useful for production agent endpoints on Azure.

1. Model Context Protocol (MCP):

• A standard way for agents to discover and call tools and data sources.
• Keeps tool integrations clean and reusable across agents.

Models and RAG Stack

1. Azure OpenAI:

• Host GPT-class models for planning, tool calling, and responses.
• Use with enterprise security and regional controls.

1. RAG for Agents:

• Give agents fresh knowledge from your docs, tickets, and databases.
• Retrieve, ground answers, and reduce hallucinations.

1. Vector Search and Indexes:

• Store embeddings and retrieve the right chunks for each agent step.
• Works with Azure AI Search and similar services.

1. Function / Tool Calling:

• Let the model choose structured tools with clear inputs and outputs.
• Add validation, retries, and timeouts around every tool.

Evaluation and Safety

1. Agent Evaluation Harness:

• Test agents on real tasks and score success, cost, and latency.
• Catch regressions before you ship a new prompt or tool.

1. Guardrails and Policies:

• Block unsafe actions and require approval for high-risk steps.
• Keep agents inside clear business rules.

1. Tracing and Observability:

• Log prompts, tool calls, and outcomes for debugging.
• Measure quality and cost in production.

Security and Identity

1. Microsoft Entra ID:

• Give each agent a clear identity and least-privilege access.
• Control who can run which tools.

1. Secrets and Key Vault:

• Store API keys and credentials outside prompts and code.
• Rotate secrets and audit access.

Cost and Routing

1. Model Routing:

• Send simple steps to smaller models and hard steps to stronger ones.
• Cut cost while keeping quality where it matters.

1. Budget Controls:

• Set token budgets, rate limits, and stop rules per agent.
• Alert when a run is too expensive or too long.

Azure Hosting

1. Azure Container Apps / AKS:

• Host agent APIs and workers with autoscaling.
• Run long tasks as background jobs when needed.

1. Vector Search and Indexes:

• Microsoft Foundry and Azure AI services help you manage models and agents in one place.
• Add monitoring, alerts, and safe rollouts.

Developer Toolkit

1. Python + REST APIs:

• Build tools agents can call and wrap your business systems.
• Prototype agents quickly, then harden for production.

1. LangChain / LangGraph:

• Compose chains, graphs, memory, and tools for agent apps.
• Move from a demo notebook to a maintainable agent service.

1. Azure SDKs:

• Use Azure SDKs for OpenAI, Search, Key Vault, and identity.
• Wire agents into your cloud with secure defaults.

Download course here

Course Modules

Part 1: Introduction to Agentic AI

Module 1: Overview of Agentic AI

• What AI Agents Are
• From Chatbots to Goal-Driven Agents
• Where Agents Help in Business

Module 2: LLM Basics for Agents

• Prompts, Tokens, and Context Windows
• When to Use an Agent vs a Simple Prompt
• Overview of Azure OpenAI and APIs

Part 2: Core Agent Building Blocks

Module 3: Agent Loops

• Plan, Act, Observe, and Repeat
• Goals, Constraints, and Stop Conditions
• Measuring Task Success

Module 4: Tool Calling

• Function Schemas and Structured Outputs
• Retries, Timeouts, and Error Handling
• Safe Tool Design and Validation
• Hands-On Tool Calling Labs

Module 5: Planning and Memory

• Short-Term and Long-Term Memory
• Planning Strategies and Scratchpads
• Memory Stores and Retrieval

Module 6: Model Context Protocol (MCP)

• What MCP Is and Why It Matters
• Exposing Tools and Data through MCP
• Hands-On MCP Connections

Part 3: Multi-Agent Orchestration

Module 7: Multi-Agent Patterns

• Planner, Researcher, and Executor Roles
• Handoffs and Shared Memory
• Human-in-the-Loop Approvals

Module 8: LangGraph

• Building Stateful Agent Graphs
• Branching, Loops, and Checkpoints
• Hands-On LangGraph Workflows

Module 9: Microsoft Agent Framework

• Building Enterprise Agents with Microsoft Tools
• Identity and Policy Integration
• Hands-On Microsoft Agent Labs

Module 10: Microsoft Foundry Agent Service

• Hosting and Operating Agents in Foundry
• Endpoints, Scaling, and Lifecycle
• Hands-On Foundry Agent Service

Part 4: RAG, Evaluation, and Security

Module 11: RAG for Agents

• Grounding Agents with Your Data
• Chunking, Retrieval, and Citations
• Agentic RAG Patterns

Module 12: Evaluation and Guardrails

• Offline and Online Evaluation
• Safety Filters and Approval Gates
• Red-Teaming Agent Behaviors

Part 5: Cost, Identity, and Azure Deploy

Module 13: Cost and Model Routing

• Choosing the Right Model per Step
• Budgets, Caching, and Rate Limits
• Hands-On Cost Control Labs

Module 14: Security and Identity for Agents

• Entra ID, Roles, and Least Privilege
• Secrets, Audit Logs, and Data Boundaries
• Threats Unique to Agents

Module 15: Deployment on Azure

• Container Apps, Functions, and APIs
• Monitoring, Alerts, and Rollbacks
• Production Checklists for Agents

Part 6: Capstone and Certification

Module 16: Capstone Project Studio

• Pick a Real Business Problem
• Design Tools, Memory, and Guardrails
• Build and Demo Your Agent

Module 17: Ship on Azure

• Deploy the Agent Service
• Add Identity, Logging, and Alerts
• Operate and Improve After Launch

Capstone Project

• Designing and Building an Agentic AI Application
• End-to-End Development: Tools, Orchestration, RAG, and Azure Deploy
• Demonstrating a Multi-Agent Solution to a Real Business Problem

Assessment and Certification

• Quizzes and Assignments
• Practical Lab Projects
• Final Project Presentation and Evaluation
• Certification of Completion