ChatGPT-powered assistant for a multi-language document base

Challenge

A Swiss company maintained a large, multi-language document collection (German, French, Italian, English). Staff lost significant time on manual search and routinely pulled in subject-matter experts for routine questions.

Approach

Built a Blazor front end on .NET, integrating Microsoft Semantic Kernel to orchestrate GPT-based retrieval-augmented generation over documents indexed in SQLite. Shipped through Azure DevOps pipelines.

Outcome

Staff get instant, language-independent answers grounded in the real document set. Manual search time dropped noticeably, freeing up experts from routine questions.

// Tech stack

• SQLite
• .NET
• Blazor
• Semantic Kernel
• Azure DevOps