Decoding AI Agents with CrewAI by Ishan Dutta

Description

🎙️ Join Ishan Dutta, Machine Learning Engineer at Adobe and formerly NVIDIA, as he breaks down AI agents from first principles and shows you how to build them using CrewAI. Most people confuse LLMs with AI agents, but understanding the distinction is crucial for building effective systems. An LLM is like a brain—it can process and reason—but an agent is like a complete human that can use tools, remember context, and execute tasks in the real world. In this workshop, you'll learn the core concepts behind agentic AI and implement working systems step by step. What you'll learn 🔹 Understand the fundamentals of AI agents and how they differ from LLMs—think of LLMs as the brain and agents as the complete human that can take action. 🔹 Learn the CrewAI framework to build single and multi-agent systems with tools, memory, and specialized tasks. 🔹 Build practical AI agents including web scrapers, PDF summarizers, and a DevOps troubleshooting system that analyzes logs, searches for solutions, and provides actionable fixes. 00:00:00 - Intro: From Hackathons to Adobe Acquisition 00:03:18 - How I Landed an NVIDIA Internship via LinkedIn 00:08:38 - LLMs Explained Simply (No Jargon) 00:13:25 - AI Agents 101: Think of Them as Humans 00:16:56 - Agent Architecture: Brain, Tools, Tasks 00:22:44 - Agent vs Tools: The Crucial Misconception 00:27:32 - The Missing Piece: Agent Memory 00:29:12 - Why CrewAI: Teams of Specialized Agents 00:30:19 - CrewAI Setup: Tools, Models, Temperature 00:34:28 - Define Your Agent: Role, Goal, Backstory 00:37:01 - Tasks & Crew Kickoff: Run Your Agent 00:40:15 - Live Demo: Web Scraper Agent 00:43:34 - Live Demo: PDF Summarizer Agent 00:50:10 - Build a Multi-Agent DevOps Debugger 00:57:30 - Debugger in Action: Analyze, Search, Fix 00:59:33 - Explore Memory: Short-Term vs Long-Term 01:01:00 - Free GitHub Lab: Beginner to Advanced Agents