Text-to-SQL Analytics System

Text-to-SQL Analytics System is an LLM-powered analytics system that translates natural language questions into structured SQL queries, executes them against relational data, and returns usable business insights without requiring manual query writing.

Key results

• Reduced SQL query turnaround time by approximately 70% for analytics-style questions.
• Enabled self-serve access to structured data insights across real business-style query workflows.
• Built a natural-language-to-database pipeline with query generation, execution, and result presentation.

What I built

• Built a natural language interface for generating SQL from user questions over structured relational data.
• Added schema-aware prompting and query validation to improve correctness and reduce broken generations.
• Designed the pipeline to execute generated SQL, return results, and support analytics-oriented workflows.
• Focused on making database access more usable for non-technical users while preserving the structure of underlying data systems.

Why this matters

Most analytics systems still depend on someone writing SQL by hand. This project turns natural language into structured database interaction, making data access faster and more usable without removing the rigor of relational systems.