Hybrid RAG with Reranker

Retrieval-Augmented Generation (RAG) systems are powerful, but their effectiveness depends on how well relevant information is retrieved. Traditional approaches often struggle to balance semantic understanding and keyword precision , which can impact answer quality.

In this session, we will explore Hybrid RAG, combining keyword-based and vector-based retrieval to improve grounding in real-world data. We’ll also look at the role of rerankers , which help refine retrieved results and enhance overall relevance and accuracy.

Through a practical walkthrough, you’ll understand how RAG pipelines are structured, evaluated, and optimized for real-world applications.

This is a practical and system-focused session designed to help build more accurate and reliable RAG systems.

Key Takeaways:

• Hybrid Search – combining keyword and vector retrieval
• RAG Pipeline – understanding end-to-end architecture
• Rerankers – improving retrieval precision
• Evaluation – key metrics like MRR and HitRate
• Production Insights – observability and common failure modes
• Practical Demo – walkthrough of a working system

Ravindra Babu Katiki is a Strategic Technology Advisor at Vriksha.ai, focusing on platform reliability and applied AI.

With extensive experience in large-scale systems, he has worked on high-scale platforms and builds AI-driven solutions across domains. His areas of interest include Hybrid RAG, Conversational AI, multi-agent systems, and observability.