Skip to main content
All use cases

Text-to-SQL Over 1,200-Table Data Warehouse

2-day turnaround → seconds

Client Context

A financial services company with a 1,200-table data warehouse containing 8 years of transaction, customer, and operational data. Business users depend on a small analytics team for every data request.

The Challenge

Analysts have a 2-day turnaround on data requests. Business users can't access their own data without SQL expertise. The analytics team is a bottleneck, and by the time answers arrive, the questions have often changed.

Our Approach

We build a Text-to-SQL pipeline with a semantic schema catalogue that vectorises every table and column description into Qdrant using BGE-M3 embeddings. When a user asks a question, the system first identifies relevant tables, then generates accurate SQL, validates it, and returns results in natural language with the underlying query visible.

Timeline: 14 weeks

The Results

  • Queries that previously required a 2-day analyst turnaround return in seconds
  • Business users projected to self-serve 70% of data requests
  • Analytics team freed for complex analysis and model development
  • 1,200 tables catalogued with semantic descriptions

Facing a similar challenge?

Let’s talk about how we can help your organisation achieve similar results.