Traditional keyword search often falls short when users phrase queries in natural language or search for concepts rather than exact terms. This talk explores how to enhance OpenSearch with machine learning techniques to deliver smarter, context-aware results. By combining semantic models like BERT or sentence-transformers with OpenSearch’s k-NN and hybrid search capabilities, we can implement semantic search that understands meaning, not just keywords. Attendees will learn how to index and search dense vector embeddings, integrate ML inference pipelines, and balance precision with recall using hybrid scoring strategies. The session will include real-world examples, deployment patterns, and performance tips to help teams bring intelligent search to their applications.