Skip to content

Mohitfegade/Fairsight

Repository files navigation

FairSight — AI Bias Detection Platform

"Catch the bias. Before it catches someone."

Live Demo Solution Challenge 2026 Firebase Gemini AI

FairSight detects hidden discrimination in AI decision systems — in seconds, for free, with zero technical knowledge required.


The Problem

Every day, AI systems make life-changing decisions about who gets a loan, a job, or medical care. These systems learn from historical data — data that reflects decades of human bias. The AI learns the bias. The bias gets automated. Real people get hurt.

Real documented cases:

  • Amazon (2018) — AI hiring tool penalized resumes containing the word "women's"
  • COMPAS (2016) — Court sentencing algorithm rated Black defendants 2x as likely to reoffend
  • US Hospitals (2019) — Healthcare AI gave less care priority to Black patients
  • Indian Banks (2023) — Credit scoring AI systematically rejected rural applicants

What FairSight Does

Upload any CSV dataset → get a fairness score in seconds → download a compliance PDF.

Core Features

  • Fairness Score 0-100 — Overall bias rating with color-coded risk level
  • 4 Fairness Metrics — Demographic parity, statistical parity ratio, 4/5ths rule, high-risk groups
  • Explainable Drill-Down — Click any demographic group, Gemini explains exactly why they face bias
  • 6-Regulation Compliance Mapper — EU AI Act 2024, India DPDP Act 2023, US EEOC, RBI Fair Practices
  • Score Improvement Simulator — Toggle fixes on/off and watch the projected score update in real time
  • Multi-Dataset Comparison — Upload before and after datasets, see side-by-side improvement
  • Google Sheets Integration — Import data directly from Google Sheets, no CSV download needed
  • Professional PDF Report — 4-page compliance document with unique audit ID, regulatory citations
  • Audit History — All audits saved to Firebase Firestore, track improvement over time
  • Intersectional Bias Detection — Detects bias across combined demographic groups

Tech Stack

Technology Usage
Google Gemini 2.0 Flash Explains bias in plain English, generates fix recommendations
Firebase Auth Google Sign-In for user accounts
Firebase Firestore Stores audit history per user
Firebase Hosting Live deployment at fairsight-26f55.web.app
React Frontend framework
Recharts Bias visualization charts
Papa Parse Client-side CSV parsing — data never leaves browser
jsPDF Professional compliance report generation

How to Run Locally

git clone https://github.com/YOUR_USERNAME/fairsight.git
cd fairsight
npm install

Create a .env file: REACT_APP_GEMINI_KEY=your_gemini_api_key REACT_APP_FIREBASE_KEY=your_firebase_key REACT_APP_FIREBASE_AUTH=your_auth_domain REACT_APP_FIREBASE_PROJECT=your_project_id

npm start

Demo Datasets

FairSight works with any CSV. Try these real-world bias datasets:

Dataset Bias Type Source
COMPAS Recidivism Racial bias in criminal justice ProPublica
Adult Income Gender pay gap UCI ML Repository
German Credit Age and gender bias in lending UCI ML Repository

Privacy

All CSV processing happens entirely in the browser. Your data never reaches our servers. FairSight uses Papa Parse for client-side parsing and processes everything in JavaScript memory. Only anonymized audit scores are saved to Firebase (when signed in).


Regulatory Coverage

FairSight checks compliance against:

  • EU AI Act 2024 — Article 10 (Data governance) and Article 13 (Transparency)
  • India DPDP Act 2023 — Section 4 (Lawful processing) and Section 6 (Consent)
  • US EEOC Guidelines — 4/5ths rule for employment selection
  • RBI Fair Practices Code — Section 3 (Non-discriminatory lending)

Built For

Solution Challenge 2026 India — Unbiased AI Decision Track

Powered by Google Gemini API and Firebase

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages