Explore the journey of the MongoDB AI assistant, from its origins as a "Skunk Works" hackathon prototype to a robust production tool integrated across the MongoDB ecosystem. Product Designer Liz Hill and AI Engineer Ben Perlmutter discuss the technical architecture and design philosophies required to build reliable, agentic AI experiences. Learn how MongoDB utilizes Atlas Vector Search and Voyage 4 embedding models to implement Retrieval-Augmented Generation (RAG), achieving a 10% increase in response quality. The discussion covers the critical transition from "vibes-based" prototyping to evaluation-driven development, ensuring data security, and the future of autonomous AI agents in developer workflows. 00:00:00 Introduction to the MongoDB AI Assistant team 00:00:21 From Hackathon prototype to production tool 00:01:10 Leveraging Atlas Vector Search for AI workloads 00:02:11 Expanding the assistant across the MongoDB ecosystem 00:03:05 Technical Architecture: How the AI agent processes queries 00:03:59 Improving quality with Voyage 4 embedding models 00:04:54 Design Philosophy: Building trust and user control 00:06:15 Solving the challenges of probabilistic systems 00:08:10 Evaluation-driven development vs. "Vibes" 00:10:12 The future of autonomous agentic actions Subscribe to the MongoDB for Developers YouTube Channel: https://www.youtube.com/@MongoDBDevelopers?sub_confirmation=1 Sign-up for a free cluster → https://www.mongodb.com/cloud/atlas/register Subscribe to MongoDB YouTube→ https://mdb.link/subscribe Visit Mongodb.com → https://mdb.link/MongoDB Read the MongoDB Blog → https://mdb.link/Blog Read the Developer Blog → https://mdb.link/developerblog MongoDB for Developers YouTube Channel → https://www.youtube.com/@MongoDBDevelopers

 276      9