Case study / Build track

ClarityCheck

A PDF trust scanner that detects hidden text, metadata issues, and manipulation attempts before people or AI systems rely on a document.

Background / Architecture

What this system proves

Built a React and FastAPI app that lets users upload a PDF, review hidden-content findings, and download a sanitized version.

Engineered detector modules for invisible text, off-page content, tiny text, low-contrast content, annotations, metadata, and zero-width characters.

Added optional Gemini-powered semantic refinement to help distinguish harmless hidden content from AI prompt-injection traps.