🔍 FairLens — AI Bias Auditing Platform

Making AI Fair for Everyone
Detect, measure, and fix hidden bias in AI systems — powered by Google Gemini

---

🎯 The Problem

AI systems used in hiring, lending, criminal justice, and healthcare are making biased decisions that discriminate against real people — and most organizations don't even know it.

Real-World Case	What Happened
COMPAS (US Courts)	Black defendants rated HIGH risk at 2× the rate of white defendants
Amazon Hiring AI	Algorithm penalized resumes containing "women's" — 57% score penalty
Healthcare AI	Black patients 41% less likely to be correctly identified as high-need

83% of HR AI tools show gender bias. These aren't hypothetical — these are documented cases.

---

💡 Our Solution

FairLens is a full-stack AI-powered platform that makes bias auditing accessible to everyone — no ML expertise required.

Upload CSV → 4 Fairness Metrics in 60s → Gemini AI Explains in Plain English → PDF Report

✨ Key Features

Feature	Description
🔍 One-Click Bias Audit	Upload any CSV → get DPD, EOD, DIR, Cramér's V metrics instantly
🤖 Gemini AI Explanations	Plain-English verdict, key findings, who's affected, fix suggestions
⚖️ Legal Compliance	Auto-checks against EU AI Act, EEOC 4/5 Rule, ISO 42001
📄 PDF Audit Reports	Professional downloadable reports for compliance teams
🌐 10 Language Support	Re-generate analysis in Hindi, Spanish, Arabic, French & more
💬 Ask Gemini Q&A	Interactive chat — ask questions about your audit results
🎯 What-If Simulator	Remove attributes & see fairness score change in real-time
🔗 Shareable Audit Links	Make any audit public with one click — no login needed to view
⚡ No-Login Live Demo	Try 3 real datasets instantly — COMPAS, Adult Income, German Credit
🔐 Google Sign-In	One-click OAuth 2.0 authentication
🌗 Dark/Light Theme	Full theme support across the dashboard
📱 Mobile Responsive	Works seamlessly on desktop, tablet, and mobile

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                    FRONTEND (Next.js 16)                     │
│  Landing · Dashboard · Audit Results · Demo · Auth Pages     │
│  Tailwind CSS · Framer Motion · Recharts                     │
│                    Deployed on Render                         │
└──────────────────────┬──────────────────────────────────────┘
                       │ REST API (Axios)
┌──────────────────────▼──────────────────────────────────────┐
│                   BACKEND (FastAPI Python)                    │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌─────────────┐ │
│  │Auth      │  │Audit     │  │Report    │  │Gemini       │ │
│  │Router    │  │Router    │  │Router    │  │Service      │ │
│  │JWT+OAuth │  │Bias Eng. │  │PDF Gen.  │  │Multi-Key    │ │
│  └──────────┘  └──────────┘  └──────────┘  └─────────────┘ │
│                    Deployed on Render                         │
└───────┬─────────────────────────────────┬───────────────────┘
        │                                 │
┌───────▼───────────┐          ┌──────────▼──────────────────┐
│  Neon PostgreSQL  │          │  Google Gemini 2.5 Flash    │
│  (Cloud Database) │          │  (AI Analysis + Multi-Key)  │
└───────────────────┘          └─────────────────────────────┘

---

🛠️ Tech Stack

Layer	Technology
Frontend	Next.js 16 + TypeScript + Tailwind CSS
Animations	Framer Motion
Charts	Recharts
Backend	FastAPI (Python 3.11)
AI Service	Google Gemini 2.5 Flash (multi-key rotation)
Bias Analysis	fairlearn + pandas + scikit-learn + scipy
Auth	JWT (python-jose + bcrypt) + Google OAuth 2.0
Database	Neon PostgreSQL (Cloud) / SQLite (local)
PDF Reports	ReportLab
Deployment	Render (Singapore region)

Google Technologies Used

• ✅ Google Gemini 2.5 Flash — AI-powered bias explanations, fix suggestions, Q&A chat, multi-language support
• ✅ Google Sign-In (OAuth 2.0) — One-click authentication
• ✅ Cloud Deployment — Fully deployed and live on Render

---

📊 Fairness Metrics Explained

Metric	Fair Threshold	What It Measures
Demographic Parity Diff. (DPD)	< 0.10	Gap in positive decision rates between groups
Equalized Odds Diff. (EOD)	< 0.10	Gap in error rates (false positives/negatives) between groups
Disparate Impact Ratio (DIR)	> 0.80	80% rule — EEOC 4/5 Rule from US employment law
Cramér's V (Bias Attribution)	—	How strongly each attribute drives unequal outcomes

---

⚡ Quick Start (Local Development)

Prerequisites

• Python 3.11+
• Node.js 18+
• Gemini API key from aistudio.google.com

1. Clone & Setup Backend

git clone https://github.com/Manin1311/Exceptional_Duo.git
cd Exceptional_Duo/backend
pip install -r requirements.txt
copy .env.example .env        # Windows
# cp .env.example .env        # Mac/Linux

2. Configure API Keys

Edit backend/.env:

GEMINI_API_KEY=your_primary_key_here
GEMINI_API_KEY_1=your_second_key_here   # optional, for rotation
SECRET_KEY=any-long-random-string
DATABASE_URL=sqlite:///./fairlens.db
ALLOWED_ORIGINS=http://localhost:3000

💡 Multiple API keys enable automatic rotation when one hits rate limits.

3. Start Backend

uvicorn main:app --reload --port 8000

4. Start Frontend

cd ../frontend
npm install
npm run dev

5. Open in Browser

URL	Page
http://localhost:3000	Landing page
http://localhost:3000/demo	Live demo (no login needed)
http://localhost:3000/register	Create account
http://localhost:3000/dashboard	Audit dashboard

---

---

📁 Project Structure

Exceptional_Duo/
├── frontend/                  # Next.js 16 + TypeScript
│   ├── app/
│   │   ├── page.tsx           # Landing page
│   │   ├── login/             # Login page
│   │   ├── register/          # Registration page
│   │   ├── dashboard/         # Audit dashboard
│   │   ├── demo/              # No-login demo page
│   │   └── audit/
│   │       ├── new/           # New audit wizard
│   │       ├── [id]/          # Audit results (authenticated)
│   │       └── public/[id]/   # Shared public audit view
│   └── lib/
│       ├── api.ts             # Axios API client
│       ├── auth-context.tsx   # Auth provider
│       └── wake-up.tsx        # Backend wake-up ping
├── backend/                   # FastAPI Python
│   ├── main.py                # App entry point
│   ├── Dockerfile             # Docker build
│   ├── requirements.txt       # Python dependencies
│   ├── routers/
│   │   ├── auth.py            # JWT + Google OAuth
│   │   ├── audit.py           # Bias analysis + demo + share
│   │   └── report.py          # PDF report generation
│   ├── services/
│   │   ├── gemini_service.py  # Gemini AI (multi-key rotation)
│   │   └── bias_engine.py     # Fairness metrics engine
│   ├── models/
│   │   ├── database.py        # SQLAlchemy models
│   │   └── schemas.py         # Pydantic schemas
│   └── demo_datasets/         # Pre-loaded CSV datasets
│       ├── compas.csv
│       ├── adult_income.csv
│       └── german_credit.csv
└── render.yaml                # Render deployment blueprint

---

🔮 Future Roadmap

• 🔄 Auto-Mitigation — Automatically rebalance datasets to reduce bias
• 📊 CI/CD API — REST API for pipeline integration (audit before deploy)
• 👥 Team Workspaces — Multi-user organizations with role-based access
• 📈 Bias Monitoring — Track fairness scores over time
• 🧠 NLP & Image Bias — Extend beyond tabular data
• 🏛️ More Regulations — India DPDP Act, Canada AIDA

---