Ajay Rajendra Kumar

I am a Computational Research Specialist at the University of Michigan, working with Prof. Shinobu Kitayama in the Culture and Cognition Lab. My primary research interests lie in multimodal learning, particularly building AI systems that can reliably reason over visual, temporal, and contextual information. I am also interested in exploring how advances in multimodal reasoning can be applied to cognitive and social science domains.

Email  /  CV  /  Scholar  /  Github  /  LinkedIn

Research

My primary research focuses on multimodal reasoning, with an emphasis on long-horizon video understanding, evidence grounding, and reliable reasoning. I am particularly interested in how multimodal models can identify, retrieve, and integrate context-relevant evidence to reason over spatial, temporal, and causal relationships.

At Prof. Lorenzo Torresani's Lab, I worked on video understanding and multimodal reasoning with vision-language models. Previously, at Prof. Huaizu Jiang's Visual Intelligence Lab, I worked on efficient visual SLAM, developing dense visual perception methods for real-time deployment on edge devices.

Currently, at the University of Michigan, I work with Prof. Shinobu Kitayama on computational models of social behavior, cooperation, and the emergence of social hierarchies. This experience has also broadened my interest in exploring applications of multimodal reasoning in cognitive and social sciences, particularly for understanding complex human interactions.

A dense visual SLAM system designed specifically for stereo and RGB-D sensors, optimized to run real-time on resource-constrained edge devices.

Combines transformer-based language models with user association dynamics to detect and counter vaccine misinformation on social media platforms.

Design and source code from Jon Barron's website