Inspiration

Recent abuse of AI for policing.

Solution

Zebra.AI aims to use stratified 2 sample t-tests and autoencoders to generate unique testing data from race segmented data points. We use these to run racial bias detection in classification models.

Vision

We aim to expand Zebra.AI to other types of models and to develop an educational interpretability dashboard to help ML engineers understand the flaws in their models.

Share this project:

Updates

Submission history