Ep064: Agentic Gen AI Experiences with Atlas Vector Search and Amazon Bedrock

Benjamin Flast, Director, Product Management at MongoDB discusses vector search capabilities, integration with AWS Bedrock, and its transformative role in enabling scalable, efficient, and AI-powered solutions.

Topics Include:

• Introduction to MongoDB's vector search and AWS Bedrock
• Core concepts of vectors and embeddings explained
• High-dimensional space and vector similarity overview
• Embedding model use in vector creation
• Importance of distance functions in vector relations
• Vector search uses k-nearest neighbor algorithm
• Euclidean, Cosine, and Dot Product similarity functions
• Applications for different similarity functions discussed
• Large language models and vector search explained
• Introduction to retrieval-augmented generation (RAG)
• Combining external data with LLMs in RAG
• MongoDB's document model for flexible data storage
• MongoDB Atlas platform capabilities overview
• Unified interface for MongoDB document model
• Approximate nearest neighbor search for efficiency
• Vector indexing in MongoDB for fast querying
• Search nodes for scalable vector search processing
• MongoDB AI integrations with third-party libraries
• Semantic caching for efficient response retrieval
• MongoDB's private link support on AWS Bedrock
• Future potential of vector search and RAG applications
• Example use case: Metaphor Data's data catalog
• Example use case: Okta's conversational interface
• Example use case: Delivery Hero product recommendations
• Final takeaways on MongoDB Atlas vector search

Participants:

• Benjamin Flast - Director, Product Management, MongoDB

See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon/isv/