This talk will enable the audience to confidently choose, design, and implement the right retrieval architecture—RAG, ReFrAG, or agentic retrieval—based on their problem, data, and production constraints. Attendees will learn how to move beyond basic RAG setups by understanding when feedback-driven retrieval (ReFrAG) is necessary and how agentic workflows can orchestrate retrieval, evaluation, and decision-making at scale.

By the end of the session, participants will be able to:

Identify common failure modes of traditional RAG systems

Decide when ReFrAG or agentic retrieval patterns provide measurable improvements

Design retrieval pipelines that incorporate feedback loops, evaluation layers, and guardrails

Apply practical architectural patterns that improve reliability, accuracy, and cost control in production LLM systems

The goal is to equip the audience with actionable mental models and design patterns they can immediately apply to build AI systems that are more trustworthy, scalable, and production-ready—rather than relying on trial-and-error or demo-only approaches.