Challenge: An operations team spent hours hunting through contracts and reports for specific clauses and figures.

What we built: An ingestion pipeline that OCRs and chunks documents, extracts structured fields, and answers natural-language questions with page-level citations. Role-based access keeps sensitive documents scoped to the right teams.

Impact: Lookups that took 20 minutes now take seconds, with an auditable trail back to the source page.

Stack: Claude API, OCR, vector store, Laravel, RBAC middleware, MySQL full-text.