cv

Doctor of Philosophy (PhD) in Computer Engineering

• Advisor: Dr. Kaushik Roy
• Research: Cognitive Inspired AI, Scene Understanding, Robust Deep Learning, Multimodal Models

MS in Computer Engineering

• Advisors: Dr. Chinmay Hegde & Dr. Soumik Sarkar
• Research Area: Domain Adaptation, Autonomous Driving, Generative Adversarial Networks, Robust Deep Learning

Graduate Research Assistant

• Foveated Object-Centric Learning (FocL)
  • Designed a multi-viewpoint foveated-vision training framework improving adversarial robustness by 61% with 56% less labeled data.
  • Integrated SAM-based dorsal-ventral pipeline, boosting ImageNet-V2 Top-1 by +7 pp and COCO zero-shot mAP by 3-4%.
• Multimodal Privacy
  • Extending framework to Self-Supervised Learning (SSL) for ViT pretraining and VLM grounding, targeting reduction of memorization in foundation models.
• Active Vision for Enhanced Robustness
  • Developed a dorsal-ventral active-vision system with ViT ventral backbones, improving robustness to state-of-the-art transformer-based black-box transfer attacks.
• Adaptive Foveated Vision & FastVLM Integration
  • Developing a token-efficient adaptive-resolution framework using factorized positional embeddings and Patch-n-Pack to process native-resolution object crops.

Graduate Research Assistant

• Attribute GAN-Based Semantic Adversarial Attacks
  • Developed a conditional GAN-based optimization scheme to generate semantic adversarial examples, driving target classifier accuracy down to 1% on CelebA.
• GAN-Based Domain Adaptation for Autonomous Driving
  • Built an attribute-controlled GAN to synthesize adverse driving conditions (e.g. rain, night), improving detector mAP by 3-5% on BDD and KITTI.

Video Research Intern

• Joint Learning of Optical Flow, Depth and Camera Pose via Transformers
  • Developed a multi-decoder Transformer architecture for autonomous-driving videos in the Multimedia R&D team.
  • Designed a self-supervised training framework leveraging epipolar geometry and view synthesis to learn flow, depth, and pose jointly in a resource-efficient manner for edge deployment.

Reviewer

• Computer Vision: CVPR (2023, 2024, 2026), ICCV (2023), ECCV (2024)
• AI: ICLR (2025, 2026), ICML (2025), NeurIPS (2025), AAAI (2026), TinyICLR (2024)
• Journals: Transactions on Machine Learning Research (TMLR), September 2025-Present

Workshop Organizer/Reviewer

• Machine Learning for Autonomous Driving Workshop

Career Chair