As AI agents demand dynamic, multi-step reasoning over enterprise knowledge,
OpenSearch has evolved beyond keyword search into the backbone of Agentic RAG
(Retrieval-Augmented Generation) systems.

This session shares real-world experience operating OpenSearch at Nagarro to power
AI agents that don't just retrieve — they reason, re-rank, and iteratively query to
deliver accurate, context-aware answers.

What you'll learn:

- What makes RAG "agentic" — multi-hop retrieval, tool-calling agents, and
iterative query reformulation
- Configuring k-NN indexes (HNSW vs. IVF) for low-latency agent response loops
- Hybrid search (BM25 + semantic) to maximize retrieval accuracy per agent step
- Ingest pipelines for chunking, embedding, and metadata enrichment at scale
- Re-ranking with OpenSearch ML models to improve agent context quality
- Monitoring and observability: tracking retrieval quality in agentic workflows
- Hard lessons: failure modes in production agentic RAG and how we fixed them

Attendees leave with a practical blueprint to operate OpenSearch as the retrieval
engine for AI agents — applicable to chatbots, copilots, and enterprise
knowledge assistants.