I am a second-year Computer Sceince Ph.D. student at the University of Maryland, advised by Soheil Feizi.
My research focuses on Image Generative Models and Vision-Language Models, with a particular interest in understanding and interpreting how these models function. I am especially interested in enhancing compositional abilities, accelerating inference, localizing and editing knowledge, unlearning undesirable information, and improving image editing capabilities in text-to-image generative models.
In the past, my research has spanned various domains, including Deep Learning Robustness, 3D Vision, and the application of Machine Learning to address neuroscientific challenges, such as improving the performance of seizure detection devices.
Email / Google Scholar / Linkedin / Github / CV
Our paper presents a method to localize where knowledge is encoded within Diffusion Transformers (DiT), enabling more interpretable and efficient model editing for personalization and knowledge unlearning.
Our paper demonstrates that text-to-image generative models often fail at accurately composing attributes and relationships due to sub-optimal text conditioning by the CLIP text-encoder, and we show that significant compositional improvements can be achieved by fine-tuning a simple linear projection on CLIP's representation space.
Our approach enhances seizure detection in epilepsy patients by converting raw EEG data into informative embeddings, achieving state-of-the-art classification results with an event-based sensitivity of 100% and specificity of 99% on the CHB-MIT scalp EEG dataset.
This research offers a data-centric approach to enhance machine learning model robustness against imperceptible adversarial perturbations by detecting and removing challenging samples during training, resulting in improved adversarial training.
PhytoOracle (PO) introduces modular, scalable pipelines for processing large volumes of phenomics data, improving efficiency, enabling data fusion, and supporting multi-system trait extraction, with broad applicability across species and various data sources, including drone data.
Our study identifies vulnerabilities in current out-of-distribution (OOD) detection methods and presents the Adversarially Trained Discriminator (ATD) as a robust alternative with improved performance.