Efficient AI Architectures
Designing low-complexity deep learning models that preserve accuracy while reducing computation through transform-based layers, decorrelated attention, and state-space modeling.
Research
Efficient models for high-impact vision problems.
Medical Imaging
Developing MRI reconstruction, segmentation, and prostate cancer analysis pipelines with hybrid residual, Hadamard, and Mamba-based architectures.
Remote Sensing
Applying transfer learning and computer vision to wildfire monitoring, situational awareness, and robust environmental detection under challenging data conditions.
Hardware-Aware ML
Exploring GPU training, FPGA deployment, high-level synthesis, and model designs suited for constrained computing systems.