Summary
Subhankar Roy is an assistant professor and researcher with eight years’ experience bridging academia and industry in computer vision and deep learning, currently based in Italy. He holds a PhD (cum laude) from the University of Trento and has held research positions at Télécom Paris, FBK, Naver Labs and Aalto, focusing on domain adaptation, continual learning and open-world recognition for images, video actions and segmentation. His work emphasizes robust model adaptation under distributional and semantic shifts with limited or no supervision, and he’s recently explored leveraging large multimodal models to boost recognition. Having transitioned from early software development in enterprise Java to cutting-edge research, he combines practical engineering instincts with rigorous publication record in top-tier venues. An intriguing thread across his career is adapting methods conceived for constrained supervision to scale with modern multimodal architectures, reflecting both theoretical depth and applied sensibility.
8 years of coding experience
3 years of employment as a software developer
Doctor of Philosophy - PhD Computer Vision and Deep Learning, Doctor of Philosophy - PhD Computer Vision and Deep Learning at Università di Trento
PhD Internship, PhD Internship at Aalto University
Bachelor’s Degree Electrical and Electronics Engineering, Bachelor’s Degree Electrical and Electronics Engineering at West Bengal University of Technology, Kolkata
Erasmus, Erasmus at Chalmers University of Technology
English, Bengali, Hindi, Italian