HR Tech3 months
Algorithmic Bias Audit for Hiring Platform
Client
Enterprise Recruitment Platform
Overview
A recruitment platform needed to prove their new AI ranking system was non-discriminatory to meet new legislation.
The Challenge
Initial internal tests showed the model favored candidates from specific universities regardless of actual skill match.
Our Solution
We conducted a comprehensive bias audit using counterfactual testing and disparate impact analysis. We retrained the model using balanced datasets and implemented fairness constraints.
Tech Stack
Scikit-learnFairlearnPythonAWS SageMaker
Key Results
bias Reduction
Disparate impact ratio improved to 0.95 (compliance standard is 0.80)
legal Risk
Verified readiness for NYC AEDT Local Law 144
fairness
Demographic parity achieved across protected groups
Disparate impact ratio improved to 0.95
Fully compliant with NYC AEDT Local Law 144
Demographic parity achieved across protected groups