Inspiration: Every week, someone forwards a file or shares a document, not realizing it’s leaking personal secrets, company strategy, or risky images. We wanted a fun, instant “security checkpoint” for documents—something anyone could run before sharing, no expert needed.
What it does: RedFlag scans any document—PDFs, images, or text—and instantly reveals if it contains sensitive info, confidential data, or potential hazards. Just upload, and our AI highlights risky content, flags confidential details, and says if your file is safe to go public or should stay private.
How we built it: We combined smart multimodal AI models (Cohere, Gemini) with custom extraction tools to analyze both text and images. The frontend lets users drag-and-drop documents or paste text, with real-time results. Behind the scenes, our backend orchestrates PII checks, prompts AIs for confident classifications, and makes the whole experience simple and fast.
Challenges we ran into: We had to wrangle PDFs, images, and text all at once—and teach the AI to spot sensitive patterns no matter where they hide. Keeping answers quick and accurate, especially when AIs disagreed, meant lots of prompt engineering. Integrating feedback and human review for edge cases was tricky but crucial for trust.
Accomplishments that we're proud of: We built a real-time compliance and review tool that anyone can use, with zero setup. RedFlag’s instant warnings and clear explanations mean users actually understand document risks—no more blind sharing! Our cross-model, human-in-the-loop pipeline keeps results robust and improves as reviewers give feedback.
What we learned: Prompt design for compliance, effective multimodal extraction, and real-world edge cases were much trickier than we thought. Automating trust but allowing for expert input made our results stronger. Above all, anyone can make their document safer with a friendly, transparent check.
What's next for RedFlag? We’re rolling out automated auditing for bulk files, deeper image and text redaction, and integration with major cloud storage and messaging platforms. Soon, RedFlag will proactively warn users—not just after upload, but before risky files ever leave your device!
Log in or sign up for Devpost to join the conversation.