FairScan helps organizations detect hidden bias in their datasets and AI models before they cause real harm. Upload any CSV, select an outcome column and a protected attribute, and FairScan acts as a bias detective — computing fairness metrics, visualizing group disparities, and using Google Gemini to explain findings in plain English with actionable fix recommendations.
Built for the Google Solution Challenge 2026.
- Rapid Analysis: Upload any CSV dataset and run a full bias audit in under 30 seconds.
- Advanced Metrics: Computes Demographic Parity Difference, Disparate Impact Ratio, and Equalized Odds Difference using Microsoft Fairlearn.
- Gemini-Powered Explanations: Plain English explanations powered by Google Gemini — no data science degree required.
- Interactive Visualizations: Visual bar chart showing approval rates across demographic groups.
- EEOC Compliance: High/Medium/Low risk classification aligned with the EEOC four-fifths rule threshold.
- Exportable Reports: Generate and download a comprehensive PDF audit report.
- Instant Testing: Pre-loaded UCI Adult Income demo dataset to test the platform instantly.
- Frontend: React 18
- Backend: FastAPI (Python)
- Bias Analysis: Microsoft Fairlearn
- AI Explanation: Google Gemini API (
gemini-1.5-flash) - Deployment: Google Cloud Run + Vercel
- Startups auditing their hiring algorithms before deployment.
- NGOs checking if resource distribution is equitable across communities.
- Banks auditing loan approval models for regulatory compliance.
- Researchers studying algorithmic fairness in real-world datasets.
- SDG 10 — Reduced Inequalities
- SDG 16 — Peace, Justice and Strong Institutions
Follow these steps to run FairScan on your system:
git clone https://github.com/Cypher-redeye/FairScan.git cd FairScan
cd backend
Create virtual environment: python -m venv venv
Activate environment:
venv\Scripts\activate
Install dependencies: pip install -r requirements.txt
Create .env file and add:
GEMINI_API_KEY=your_api_key_here
Run server: python main.py
Backend will run on: http://localhost:8000
Open a new terminal:
cd frontend npm install npm run dev
Frontend will run on: http://localhost:5173
- Open frontend
- Upload CSV dataset
- View bias metrics and Gemini explanations
- Results depend heavily on dataset quality and completeness
- Gemini explanations may vary slightly due to generative nature
- Currently supports CSV datasets only
- Real-time performance may vary for very large datasets
If you are looking for the tags to add to the GitHub "About" section, here they are:
fairness, bias-detection, responsible-ai, gemini-api, google-solution-challenge, fairlearn, fastapi, react, algorithmic-fairness, ethical-ai