Natural-Language OracleDB Search

I built a natural-language-to-OracleDB search application that lets users describe what they need in plain language, then maps the request to structured database fields and search filters.

The application was designed for internal enterprise data where users often know the business question but not the exact database schema, field names, or query shape.

Highlights

• Extracted structured search intent from natural-language prompts.
• Mapped user requests to OracleDB fields and filters.
• Improved extraction quality with Qwen3:8B fine-tuning and LangExtract-style prompting.
• Improved user satisfaction by nearly 30% by reducing manual field selection and query friction.

Technologies

Python, OracleDB, local LLMs, Qwen3:8B fine-tuning, information extraction, structured search, and prompt evaluation.