Amitangshu Mukherjee

Scene Understanding · Robust Deep Learning · Multimodal Models

I’m a Ph.D. candidate in Electrical and Computer Engineering at Purdue University, working with Professor Kaushik Roy. My dissertation work establishes biologically-inspired vision systems for real-world scene understanding. My research develops adaptive active-vision frameworks to address practical challenges such as domain shift, long-tail distributions, and trade-offs between privacy and computational efficiency.
Currently, I am focusing on adaptive foveation strategies for multimodal foundation models (VLMs), integrating robustness, privacy, and efficiency to enable scalable, high-fidelity perception in complex environments.

I received my M.S. in Computer Engineering from Iowa State University, where I worked with Professor Chinmay Hegde and Professor Soumik Sarkar. During my M.S., I developed GAN-based methods for domain adaptation and adversarial attacks — targeting the robustness of perception models under domain shift, with a focus on driving-scene data for autonomous-driving applications.

Prior to that, I earned my B.Tech in Applied Electronics & Instrumentation Engineering from Heritage Institute of Technology, Kolkata (affiliated with West Bengal University of Technology).