GuardLens MVP

GuardLens is an MVP web application to monitor AI model fairness and bias.

Features

• Upload model predictions and sensitive features
• View fairness metrics: accuracy, precision, recall, selection rate
• Compute demographic parity and equalized odds
• Download a JSON fairness audit report

Setup

pip install -r requirements.txt
streamlit run app.py

Sample CSV Format

y_true,y_pred,sensitive_feature
1,1,F
0,0,M
...