OLAP for AI Applications and why you should care

RAG applications don't just need to find documents - they need to analyze them. To make this tangible, consider asking an AI real estate assistant: “Show me family-friendly houses under $800K in good school districts.”. What is expected on the output is property listings alongside market insights like price trends, days on market, neighborhood medians and school rating distributions – enabling the LLM to reason better and not just retrieve. For meaningful results, a clever combination of semantic search and analytical processing is needed. In this talk we introduce a new concept of Search-OLAP: an approach where information retrieval becomes a native analytical primitive. It is the architecture where BM25, vector similarity and SQL aggregations coexist as peers in a vectorized execution engine. Join us if you're building RAG systems, managing dual search-analytics stacks or designing applications requiring both semantic retrieval and statistical reasoning and want to learn from our mistakes and successes.