🤖 Agentic AI Patterns - DeepLearning.AI Course

Master the art of building intelligent, autonomous AI agents with Andrew Ng's comprehensive course on Agentic Design Patterns

🎯 Course Overview

This repository contains the complete coursework, lab assignments, and practical implementations from DeepLearning.AI's Agentic AI Course taught by Andrew Ng. Learn to build sophisticated AI agents that can plan, reflect, use tools, and collaborate to solve complex real-world problems.

📚 Module Structure

🔄 Module 2: Reflection Design Pattern

Master self-improving AI systems through iterative critique and refinement

• Core Concept: AI systems that critique and improve their own outputs
• Key Projects:
  • 📊 Visualization Agent: Creates and refines data visualizations through LLM feedback
  • 🃏 Flashcard Generator: Improves study materials through pedagogical reflection
  • 🔒 PII Protection: Defensive reflection for healthcare AI security
  • 💾 SQL Agent: Self-correcting database query generation

🛠️ Module 3: Tool Usage Design Pattern

Empower AI agents with external tools and function calling capabilities

• Core Concept: Extending AI capabilities through strategic tool integration
• Key Technologies: OpenAI Function Calling, AISuite, MCP (Model Context Protocol)
• Key Projects:
  • 📧 Email Agent: Complete email management with FastAPI integration
  • 🔍 Research Agent: Web search and academic research automation
  • 🗄️ SQL Agent: Database interaction and schema exploration tools

🎯 Module 4: Practical Tips for Building Agents

Production-ready strategies for robust AI agent development

• Core Concept: Real-world implementation patterns and evaluation strategies
• Focus Areas:
  • 📈 Agent Evaluation: Gold standard creation and F1-score tracking
  • 🔧 Hyperparameter Optimization: Search engine and result tuning
  • 🏗️ Production Patterns: Scalable architecture design

🤝 Module 5: Multi-Agent Collaboration

Orchestrate multiple specialized agents for complex problem solving

• Core Concept: Dividing complex tasks across specialized agent teams
• Collaboration Patterns: Sequential orchestration, concurrent dialogues, fault-tolerant automation
• Key Projects:
  • 🛍️ Customer Service Pipeline: Four-agent system (Planner → Coder → Executor → Reflector)
  • 📝 Research Team: Specialized roles for research, writing, and critique

🚀 Production-Ready Implementation

Reflective Research Agent

Full-stack FastAPI application demonstrating production agentic patterns

Tech Stack: FastAPI, PostgreSQL, Docker, Jinja2, Tavily API, arXiv API, Wikipedia API

Features:

• 🔄 Multi-step Planning: Intelligent research workflow orchestration
• 🔍 Tool Integration: Tavily search, arXiv papers, Wikipedia lookup
• 📊 Real-time Tracking: Live progress monitoring via WebSocket
• 🗄️ State Persistence: PostgreSQL for task management and results
• 🌐 Web Interface: Clean UI for research task initiation

Quick Start:

# Build and run the complete research agent
docker build -t fastapi-postgres-service .
docker run --rm -it -p 8000:8000 -p 5432:5432 --env-file .env fastapi-postgres-service

# Access the application
open http://localhost:8000

🎨 Key Workflow Visualizations

Chart Generation with Reflection

Demonstrates iterative improvement: Initial code → Execution → Critique → Refinement → Final output

Research Agent Architecture

Shows tool orchestration: User Query → Planning → Tool Selection → Execution → Report Generation

💡 Learning Outcomes

After completing this course, you'll master:

• 🧠 Agentic Design Patterns: Reflection, Tool Use, Planning, Multi-agent coordination
• 🔧 Production Implementation: FastAPI services, database integration, containerization
• 📊 Evaluation Strategies: Creating benchmarks, tracking performance metrics
• 🛡️ Security Patterns: PII protection, defensive reflection, prompt injection prevention
• 🤖 Agent Orchestration: Sequential and parallel agent coordination

🛠️ Prerequisites

• Python 3.8+
• Docker & Docker Compose
• OpenAI API Key
• Tavily API Key (for web search)

📖 Course Structure

Each module contains:

• 📹 Video Content: Comprehensive explanations of design patterns
• 🧪 Ungraded Labs: Hands-on practice with guided implementation
• ✅ Graded Labs: Assessment-ready projects with solution guides
• 📝 Documentation: Detailed README files and architectural diagrams

🎓 About the Instructor

Andrew Ng - Co-founder of Coursera, former Director of Stanford AI Lab, and pioneer in online education. This course represents the cutting-edge of agentic AI development, distilling years of research into practical, implementable patterns.

---

Ready to build the next generation of intelligent agents? Start with Module 2 and work through the progression of increasingly sophisticated agentic patterns. 🚀