Agentic RAG is all about squeezing out that last 5% of GenAI precision: why settling for 95% can land you in court, and how an LLM-as-a-Judge plus iterative feedback loops help banish hallucinations and earn real trust from users.
On the architecture side, Adi Polak dives into RAG vs. fine-tuning, hybrid retrieval hacks, and a controller-planner-judge agent stack—complete with orchestrator patterns and Kafka-powered scaling. He round-trips his talk with real SDR, marketing ops, and fraud-detection cases, then wraps up on avoiding Text-to-SQL pitfalls through smart templating.
Watch on YouTube
Top comments (0)