SPGE Enhanced Search: relevancy engine for Agentic RAG
Tomasz Florin and Artur Barczyński • Location: TUECHTIG • Back to Haystack EU 2026
As AI agents increasingly rely on enterprise search systems for retrieval, relevance becomes a critical success factor. In this session, I will present SPGE Enhanced Search, the hybrid search and relevancy platform that serves as both the primary search engine for our core enterprise platform and the retrieval layer supporting Agentic RAG workloads across hyperscaler environments. The talk focuses on how we designed and evolved a relevancy framework tailored to complex business data. I will discuss the business-driven relevance rules that shape ranking decisions, the assumptions behind our retrieval strategy, and how we combine lexical and semantic search techniques to create a hybrid search experience that balances precision, recall, explainability, and performance. The session will also cover architectural trade-offs, operational challenges, lessons learned from running search at scale, and practical guidance for teams building reliable search platforms for modern Agentic RAG applications.
Download the Slides Watch the Video
Tomasz Florin
S&P Global
Engineering leader at S&P Global Energy, where he leads the development of enterprise search and AI-powered information retrieval platform. He specializes in relevance engineering, semantic search, retrieval-augmented generation (RAG), metadata architectures, and large-scale content discovery systems. He has been instrumental in transforming traditional enterprise search into intelligent, agent-ready discovery platform that support data-driven workflows across organization. His work focuses on combining search, AI, and knowledge systems to create trusted, scalable retrieval experiences for both human users and autonomous agents
Artur Barczyński
S&P Global
Artur Barczyński is a Senior Software Engineer and Team Leader at S&P Global. He holds a Master’s in Software Engineering from the University of Oslo, and has spent much of his career on large-scale public infrastructure — including several years building systems for Norway’s national population register. Since 2018 he’s been building applications on top of different search engines, working on relevance, ranking, and hybrid search techniques (including RRF-based approaches) to help people find what they’re actually looking for. Outside of work, he’s usually off-road in his car or tinkering with retro computers.