🔗 Git Repository: github.com/NSArjun/RedactLess
📺 Video Explanation: https://youtu.be/vhkbdD8TPO8?si=pxynPT0YuRVF2jkc

Every day, organizations in healthcare, finance, and law handle massive amounts of sensitive data.
Manual redaction remains slow, inconsistent, and error-prone; even one missed field can lead to a privacy disaster ⚠️

Existing tools often:

• ❌ Miss complex or nested PII formats
• 🐢 Break data usability or structure
• 💢 Require tedious manual cleanup

That’s where RedactLess steps in:

Because Privacy Isn’t Optional.

• Frontend
  • Next.js (TypeScript)
  • Tailwind CSS
• Backend
  • Flask API
  • Python (Hugging Face Transformers, Faker)
• Environment
  • IBM Z Jupyter Notebook (AI model execution)
  • LinuxOne System (Development)

1. AI-Powered Redaction

   • Uses a fine-tuned transformer model to identify and mask Personally Identifiable Information (PII).
   • Labels entities such as FULLNAME, EMAIL, PHONE, ADDRESS, and more.

2. Sensitivity Levels

   • Light: Masks key numerical identifiers (Aadhaar, Phone, Govt IDs).
   • Medium: Adds names and addresses.
   • Heavy: Redacts every identifiable element.

3. Smart Data Synthesis

   • Extracts schema from CSV/JSON input.
   • Generates realistic synthetic data using Faker.
   • Keeps the structure intact for testing and analytics.

4. Mainframe Integration

   • Model tested and executed on IBM Z Jupyter Notebook for reliability and scalability.
   • Demonstrates seamless AI–Mainframe interoperability.

5. Web Interface

   • User-friendly web app to upload files and select sensitivity levels.
   • Built-in synthetic data generation page.

• Python 3.9+
• Node.js + npm (for the frontend)
• IBM Z Jupyter access (optional, for model execution)

git clone https://github.com/NSArjun/RedactLess
cd RedactLess/backend
pip install -r requirements.txt
python app.py

cd ../frontend
npm install
npm run dev

• The backend runs at http://localhost:5000
• The web app runs at http://localhost:3000

• All processing happens locally or on IBM Z’s secure environment.
• Built with a strong focus on compliance (GDPR, HIPAA) and data governance.

• Successfully executed and tested AI model on IBM Z Jupyter Notebook, proving mainframe–AI integration.
• Connected Flask backend to ML inference pipeline on IBM Z.
• Understood secure data flow between frontend, API, and model.
• Deployed Hugging Face Transformer for text redaction and synthesis.
• Resolved dependency, I/O, and cross-platform issues.
• Strengthened understanding of privacy-first AI and compliance in enterprise systems.

• Add support for document and image redaction (PDF, DOCX).
• Integrate Faker-based multi-language synthesis.
• Expand deployment to IBM Cloud & Z Containers.
• Build an enterprise dashboard for compliance tracking.

💬 Future Ideas

Replace stick figure with a 2D or 3D avatar

Add recording & playback features

Integrate sound-reactive effects

Stream live pose data into Unity or Blender

RedactLess is licensed under the MIT License: you’re free to use, modify, and distribute it with attribution.

Made with ❤️ by Team MAVERICK for the IBM Z Datathon

^{Arjun • Keerthi • Aravind • Shivaa • Yathin • Sowmya}