Hello! I am a 4th year Ph.D. student at Arizona State University’s School of Computing and Augmented Intelligence (SCAI). I work with Hannah Kerner on algorithmic biases in vision and language tasks. I am curious to learn what factors cause biases in AI models.

Prior to joining ASU, I worked as a Junior Research Fellow at IIIT Hyderabad with Ravi Kiran S. I worked on a variety of vision applications, including pose estimation, object tracking, and segmentation. I completed my undergraduate study in Computer Science from Amrita University in 2020.

To know more about my work and past experiences, feel free to check out my CV.

Publications

A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions

WACV 2026

Rahul Nair*, Bhanu Tokas*, Hannah Kerner

pdf

This paper introduces a novel method for measuring directional bias amplification in image captioning models. We demonstrate that bias can flow in multiple directions and propose metrics to quantify these directional effects in vision-language systems.

DPA: A one-stop metric to measure bias amplification in classification datasets

NeurIPS 2025

Bhanu Tokas*, Rahul Nair*, Hannah Kerner

pdf

We present DPA (Directional Parity Analysis), a comprehensive metric for quantifying bias amplification in classification tasks. Our approach provides a unified framework for measuring how models amplify existing dataset biases across different protected attributes.

Classification Drives Geographic Bias in Street Scene Segmentation

CVPR Workshops 2025

Rahul Nair, Bhanu Tokas, Gabriel Tseng, Esther Rolf, Hannah Kerner

pdf

This work investigates how classification-based training objectives contribute to geographic bias in semantic segmentation models. We show that models trained on street scene datasets exhibit systematic performance disparities across different geographic regions.

Metadata: a tool to supplement data science education for the first year undergraduates

ICIET 2020

Rahul Nair, Mukesh N Chugani, SK Thangavel

pdf

We introduce MetaData, an educational tool designed to enhance data science learning for first-year undergraduate students. The tool provides interactive visualizations and guided workflows to help students understand fundamental concepts in data analysis and statistics.