Check out our gen AI Showcase for gen AI code examples and apps: https://github.com/mongodb-developer/GenAI-Showcase/tree/main Check out our AI Learning Hub: https://mdb.link/Gjzf7HpsOW0-AI Subscribe to MongoDB YouTube→ https://mdb.link/subscribe The era of typing keywords and sifting through endless search results is over. We are now entering the age of Agentic Search, where AI agents search the web, synthesize information from countless sources, and deliver exactly what you need. In this episode of People Who Ship, host Apoorva sits down with Rotem Weiss, Founder and CEO of Tavily, a company building the "Internet of AI Agents". Rotem explains how AI agents are fundamentally different from human searchers, operating at a scale and speed that requires a completely new infrastructure. He dives deep into "context engineering"—the science of giving an agent just the right information, not drowning it in data. He also shares why Tavily chose MongoDB. While many new AI apps default to vector-only databases, Rotem explains why that's not the complete answer. Tavily needed the flexibility of MongoDB's hybrid search capabilities—combining semantic (vector) search, keyword search, and graph capabilities all in one platform. (00:00) What is Agentic Search? (00:30) Guest Intro: Rotem Weiss, Tavily CEO (01:10) The Origin of Tavily: GPD Researcher (01:27) The Problem: LLM Knowledge Cutoff (02:21) The Gap: Agents Need Real-Time Data (03:30) Use Cases: Fraud Detection & Personalized Sales (04:03) The Power of Research Agents (05:45) What is the "Internet of Agents"? (07:35) How AI Agents Search Differently Than Humans (08:43) The "UX" for an AI Agent (09:16) Tavily's Agent-Specific Browser (10:54) The Key is Context Engineering (11:34) How Tavily Ranks and Finds Snippets (12:04) Ranking is More Than Just Semantics (13:50) Ranking Pages vs. Ranking Inside Pages (15:09) Why Vector Databases Aren't the Whole Answer ( 15:44) Why Tavily Chose MongoDB (Hybrid Search) (16:40) The Limited Context Window Problem (18:15) Giving the Agent the Exact Right Amount of Context (20:29) What Does SEO Look Like for AI Agents? (21:19) Optimizing Content "Chunks" Semantically (23:46) The Future: Websites Written in Tokens? (24:45) The Biggest Problem in AI Agents Today (25:38) Agent Memory Should Work Like a Human Brain (27:14) Self-Improving Memory Systems (28:21) Security Risk: Prompt Injection from the Web (29:57) The Future: Balancing Latency and Accuracy (31:44) Final Advice: Master Context Engineering Visit Mongodb.com → https://mdb.link/MongoDB Read the MongoDB Blog → https://mdb.link/Blog Read the Developer Blog → https://mdb.link/developerblog

 290      11