Case Study: ICE/NYSE achieves natural language search for structured data with Databricks

ICE/NYSE Unlocking Financial Insights with a Custom Text-to-SQL Application

ICE/NYSE, a global financial organization behind the New York Stock Exchange, needed a way for non-technical users to search structured financial data in natural language without understanding schemas or writing SQL. The company partnered with Databricks and used Databricks Mosaic AI products to build a custom text-to-SQL application with retrieval-augmented generation (RAG), Vector Search, Foundation Model APIs, and Model Serving.

Databricks helped ICE/NYSE implement a closed-loop system that retrieves table metadata and sample queries, generates SQL from user questions, and continuously improves through evaluation and feedback. In just five weeks, the solution delivered 77% syntactic accuracy and 96% execution match across about 50 queries, helping ICE/NYSE turn raw financial data into actionable insights faster.

ICE/NYSE