⌘K
AI-Native TrainingModule 3
3
Associate12-15 hoursAvailable Now

Agentic AI - Building Autonomous Systems

Reflection, Tool Use, Planning, and Multi-Agent Workflows

Based on DeepLearning.AI's Agentic AI course by Andrew Ng. Build agentic design patterns that enable AI to plan, execute, reflect, and adapt to complete complex tasks autonomously.

Based on:DeepLearning.AI

Learning Objectives

Build agentic design patterns: reflection, tool use, planning, multi-agent workflows
Integrate AI with external tools: databases, APIs, web search, code execution
Evaluate and optimize AI systems for production deployment
Understand degrees of autonomy in AI agents

Topics Covered

1

Introduction to Agentic Workflows

Understanding agentic AI and its benefits

2 hours
What is Agentic AI vs traditional prompt-response
Degrees of autonomy: assisted to fully autonomous
Real-world applications and use cases
Task decomposition for agent workflows
2

The Reflection Design Pattern

AI that critiques and improves its own work

2.5 hours
Self-critique and iteration for quality improvement
Implementation: generate → critique → refine → iterate
Use cases: chart generation, SQL queries, code generation
External feedback integration
3

The Tool Use Pattern

Connecting AI to databases, APIs, and external services

3 hours
Creating tools: schemas, functions, error handling
Tool syntax standards and function calling
Model Context Protocol (MCP)
Code execution as a tool
Common tool categories: data, action, computation, search
4

The Planning Pattern

Breaking complex tasks into executable steps

2.5 hours
Upfront, iterative, hierarchical, and adaptive planning
Creating and executing LLM plans
Plan representation formats (JSON, YAML)
Planning with code execution
5

The Multi-Agent Pattern

Coordinating multiple specialized AI systems

2.5 hours
Multi-agent architectures: sequential, parallel, hierarchical, collaborative
Communication patterns: broadcast, point-to-point, pub-sub, shared state
Agent specialization: research, analysis, writing, critique, orchestrator
6

Evaluating and Optimizing Agentic Systems

Production-ready agent development

2 hours
Evaluation frameworks and metrics
Error analysis and debugging
Component-level evaluations
Cost and latency optimization
7

Agent Memory and Context Management

LLMs as Operating Systems: Managing state across interactions

2 hours
Short-term vs long-term agent memory
Context window management strategies
Semantic caching for efficiency
Memory retrieval and summarization
Persistent state across sessions

Hands-On Projects

Chart Generation Agent

intermediate3 hours

Build a chart generation agent with reflection pattern

Research Agent with Tools

intermediate3 hours

Create a web search and summarization agent

Multi-Agent Research Team

advanced4 hours

Build a market research team with specialized agents

Complete Agentic System

advanced5 hours

Implement all four design patterns in one system

Module Progress

Not started

Recommended Resources

DeepLearning.AI CoursesLangGraph Repository

External learning resources to supplement your training.

Assessment

Build complete agentic system + Deploy to production environment + Evaluation report

Module Navigation