Decoding AI Agents with CrewAI by Ishan Dutta

Описание: 🎙️ 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