How to build context-aware AI with LLMs, RAG, and MCP

Description

With all the technologies floating around the tech space these days it almost sounds like a buzzwords bingo. This talk will go over all these concepts to explain the differences and purpose of all these technologies. We’ll cover everything from explaining embeddings and different vector search strategies, to the difference between RAG and agents, what the role of MCP is, and how it all ties together. Speaker: Iulia Feroli, Senior Developer Advocate, Elastic 00:00 – Why LLMs Need Context 01:32 – What Are Embeddings and How They Work 03:15 – Dense vs Sparse Embeddings Explained 05:42 – How Context Windows Impact LLM Performance 07:10 – Vector Search and Hybrid Search Basics 09:08 – Enter RAG (Retrieval-Augmented Generation) 11:50 – Why LLMs Hallucinate and How RAG Fixes It 13:22 – From RAG to AI Agents — Automating the Pipeline 15:45 – Real-World Example: Agent Workflow with Elastic 17:40 – Introducing MCP (Model Context Protocol) 19:05 – How MCP Connects Clients, Servers and Tools