Primary case study
QueryPilot - Natural Language Database Assistant
Problem
A conversational database assistant that converts natural language questions into schema-aware SQL queries.
Approach
A schema-aware workflow uses database context to generate SQL, with human review before execution.
What I built
I designed the LangGraph workflow and built the system across a FastAPI backend, PostgreSQL schema introspection and a React frontend, with human review before query execution.
Workflow
- 01Natural Language
- 02Understand Request
- 03Schema
- 04Generate SQL
- 05Human Review
- 06Execute
Technologies
- Generative AI
- AI Agents
- LangGraph
- LangChain
- PostgreSQL
- React
- TypeScript
- FastAPI
Why it matters
The human review step keeps generated SQL visible and reviewable before it reaches the database.
