Usman Roshan

Usman Roshan is an Associate Professor in the Department of Data Science at New Jersey Institute of Technology (NJIT). He holds a Ph.D. in Computer Science from the University of Texas-Austin (2004), along with M.S. (2002) and B.S. (1998) degrees from the same institution. His research focuses on machine learning, deep learning, adversarial robustness, and medical imaging applications. Key areas include developing AI systems for surgical performance evaluation, medical image segmentation, and defense against adversarial attacks in neural networks.

Roshan's work bridges theoretical advancements and practical healthcare applications. He has pioneered AI-driven tools for real-time surgical skill assessment and low-cost surgical simulators. His recent studies emphasize adversarial machine learning, exploring robustness against black-box and white-box attacks on medical imaging models. His contributions span multi-modal deep networks for glioma survival prediction, U-Net-based vessel segmentation in carotid ultrasounds, and defense mechanisms for medical AI systems.

Publications highlight innovations in neural network architectures (e.g., cascaded systems, dual-decoder U-Nets) and novel loss functions (e.g., 01 loss) to enhance model reliability. His research also addresses challenges in bioinformatics, including genomic data analysis for cancer prediction and protein sequence classification using K-means-based feature learning.

No scientific awards are explicitly listed in the provided text. Teaching responsibilities include advanced courses in machine learning, data mining, and deep learning (e.g., DS 677-851, CS 634). His work reflects a commitment to advancing both fundamental ML theory and its clinical applications.