As a Graduate Research Assistant at Purdue University, West Lafayette (Fall 2023 – present), I develop deep learning and generative models for multimodal medical image reconstruction and restoration under complex, real-world degradations. My research emphasizes physics-informed learning for modeling motion, noise, and artifacts in high-resolution computed tomography and magnetic resonance imaging. I build physics-based motion simulation pipelines and design transformer- and generative adversarial network-based frameworks for motion and artifact correction in high-resolution computed tomography. I am also working on generalized medical image restoration under complex and mixed degradations, including motion, blur, undersampling, and domain shift, using domain adaptation, self-supervised learning, and transformer-based architectures. This includes developing dual-cycle domain-adaptive generative models and adversarial transformer frameworks for robust motion artifact correction. In magnetic resonance imaging, I have implemented the PETALUTE lung sequence at Riley Children’s Hospital and developed compressed sensing and transformer-based reconstruction methods for free-breathing lung magnetic resonance imaging, in collaboration with the University of North Carolina at Chapel Hill, the University of California, San Francisco. Additionally, I work with foundation models such as DINO and MedSAM for motion severity classification and mixed-domain semi-supervised segmentation. LinkedIn - https://www.linkedin.com/in/farhan-sadik-a37645236/ GitHub - https://github.com/fsa125