AnonymousDoc AI

This is a document processing webapp that uses intelligent redaction to protect sensitive information in PDFs.Users simply upload a document, and AI scans it for personally for sensitive information.

Comment

Inspiration

We initially considered building a playful, kid-friendly privacy tool themed around the Cookie Monster. It was fun and engaging, but ultimately we pivoted to something more universally applicable: AnonymousDoc AI. This direction allowed us to tackle real-world privacy challenges in education, healthcare, and legal contexts, where sensitive documents need to be redacted quickly and responsibly.

What It Does

AnonymousDoc AI is a privacy-first document redaction tool that:

Accepts PDF uploads via a clean, intuitive interface

Analyzes text using NLP and regex to detect sensitive entities (names, emails, locations, etc.)

Scores the document’s sensitivity level

Previews both the original and redacted versions

Generates downloadable redacted PDFs with visual overlays

It’s designed to be modular, teachable, and hackathon-ready — perfect for rapid prototyping, contributor onboarding, or CTF-style challenges.

How We Built It

Frontend: Started with vanilla HTML/CSS/JS, then migrated to React + TypeScript using Vite for modularity and maintainability

Backend: FastAPI with spaCy for entity detection, PyMuPDF for PDF parsing and redaction, and CORS middleware for seamless integration

Workflow: File upload → NLP analysis → sensitivity scoring → redaction overlay → export

We emphasized reusable logic, clean state management, and pedagogical clarity throughout.

Challenges We Ran Into

Getting PyMuPDF to reliably extract and redact text across diverse PDF formats

Handling image-based PDFs with no extractable text

Managing CORS and multipart form data in a local dev environment

Designing a scoring rubric that balances direct and indirect identifiers

Debugging async file handling between frontend and backend

Each challenge became a learning artifact — documented, modularized, and turned into flashcards or checklists for future contributors.

Accomplishments That We're Proud Of

Built a fully functional privacy tool from scratch in a short time frame with little experience

Migrated from vanilla JS to React + TypeScript with reusable components

Created a backend that not only analyzes but transforms documents

Designed a scoring system that’s explainable and extensible

Documented every major bug fix and architectural decision for teaching

What We Learned

How to bridge frontend-backend workflows with FormData and streaming responses

How to use spaCy and regex together for layered entity detection

How to scaffold modular components that are both functional and pedagogically useful

How to turn technical setbacks into reusable learning resources

How to design for trust, transparency, and privacy in UX

What's Next for AnonymousDoc AI

Expand entity detection with multilingual models and OCR for image-based PDFs

Build a contributor dashboard with flashcards, checklists, and onboarding flows

Integrate with secure cloud storage for document retention and audit trail

Expend to more file types ( etc docx)

Built With

Share this project:

Updates