Welcome page
Audit
Foresnsic audit history
Verdict + IQ + Equity — Because every AI decision deserves a fair trial.
VERDIQT is a Gemini-powered forensic AI bias auditing platform that goes beyond detecting bias. It traces exactly WHERE in the ML pipeline bias was born, WHO it harms in human terms, and HOW to fix it — with a regulator-ready audit report.
Built for the Google Solution Challenge 2026 — Build with AI Hackathon.
Computer programs now make life-changing decisions — who gets a loan, a job, or medical care. When these systems learn from flawed historical data, they silently repeat and amplify the same discriminatory patterns, harming real people without anyone noticing.
77% of AI systems deployed today have never been audited for bias.
VERDIQT gives organizations a full Bias Autopsy in under 60 seconds:
| Feature | Description |
|---|---|
| 🔢 VERDIQT Score | 0-100 fairness rating — like a credit score for your AI |
| 🔬 Forensic Bias Tracer | Traces bias across all 4 ML pipeline stages |
| 👤 Human Impact Story | Real story of the person your model is failing |
| 🔄 What-If Fix Simulator | Test fixes before applying them |
| 📄 Audit Report | Regulator-ready PDF with compliance references |
| 💚 Empathy Mode | Explains rejection to affected person in plain language |
No other tool does this. VERDIQT traces how bias traveled and amplified through all 4 pipeline stages: Data Collection → Data Labeling → Feature Selection → Model Training Each stage shows contamination %, forensic evidence, and a plain-English analogy.
Gemini generates a realistic story of a fictional person from the affected group who would be wrongly rejected. Ends with:
"This is who your model is failing."
Explains to an affected person — in plain language — why the AI rejected them unfairly, what their rights are, and what action they can take.
| Layer | Technology |
|---|---|
| Frontend | React + TypeScript + Tailwind CSS |
| AI Engine | Google Gemini 1.5 Flash API |
| Cloud Deployment | Google Firebase Hosting |
| Dataset | UCI ML Repository — Loan Prediction Dataset |
| Build Tool | Vite |
- Node.js v18+
- Google Gemini API Key (Get it here)
# Clone the repository
git clone https://github.com/yourusername/verdiqt.git
cd verdiqt
# Install dependencies
npm install
# Create environment file
cp .env.example .envAdd your Gemini API key to .env:
VITE_GEMINI_API_KEY=your_gemini_api_key_here
# Run locally
npm run devOpen http://localhost:5173
# Build the project
npm run build
# Install Firebase CLI
npm install -g firebase-tools
# Login and deploy
firebase login
firebase init hosting
firebase deployLive at: https://verdiqt.web.app
VERDIQT uses the UCI Loan Prediction Dataset (614 records) for demo:
| Group | Approval Rate |
|---|---|
| Male | 69% |
| Female | 61% |
| Graduate | 70% |
| Not Graduate | 58% |
| Urban | 76% |
| Rural | 61% |
Source: UCI ML Repository
- EU AI Act — Article 10 (Data Governance)
- India Digital Personal Data Protection Act 2023
- IEEE 7003 — Algorithmic Bias Considerations
- Real CSV file upload with auto column detection
- Support for Hiring, Medical, Education domains
- Direct ML model file upload (.pkl, .h5)
- Continuous weekly monitoring dashboard
- VERDIQT Certification Badge for compliant organizations
- Mobile app for compliance officers
Built with ❤️ for Google Solution Challenge 2026
| Role | Name |
|---|---|
| Team Lead | Aditi J N |
| Team members | Gayathri, Amrutha, Basil |
MIT License — see LICENSE for details.
"VERDIQT doesn't just tell you your AI is biased — it tells you exactly where bias was born, who it is harming right now, and gives you a fix you can apply today."
This contains everything you need to run your app locally.
View your app in AI Studio: https://ai.studio/apps/6301edcf-28fa-4d5b-abbe-9d1d42d316c5
Prerequisites: Node.js
- Install dependencies:
npm install - Set the
GEMINI_API_KEYin .env.local to your Gemini API key - Run the app:
npm run dev