Hi, I'm Ramsha! I am a Research Fellow at the University of Maryland's Institute of Health Computing, collaborating with the U.S. FDA's Division of Imaging, Diagnostics, and Software Reliability. My advisors are Professor Joseph JaJa at UM-IHC and Dr. Miguel Lago at the FDA. My research focuses on developing a robust evaluation framework for AI explainability methods in medical imaging. I am particularly interested in AI Alignment and Explainability, motivated to understand the internal workings of AI models to enable their safe and trustworthy deployment in critical domains like healthcare.
I completed my Masters in Computer Science (Machine Learning) from Georgia Institute of Technology. I have also interned at the Brookhaven National Lab working on the explainablity of Vision Transformer models.
I am happy to connect to explore any research and collaborations!
News!
- Our paper "Comparative Plausibility Evaluation of Heatmaps for Vision Transformers in Digital Mammography" has been accepted at the EXPLIMED workshop at ECAI '25. See you in Italy!
- Attending the Deep Learning for Science School (DL4Sci25) in Berkeley, CA. (June 23 - June 27, 2025).
Invited Talks and Poster Presentations
- Presented our poster "How Does Target Layer Selection Impact Class Activation Map (CAM)-based Heatmaps for Vision Transformers?" at the Poster Session at DL4Sci 25 (June 24, 2025).
- Invited Talk: Comparative Evaluation of Class Activation Map (CAM) based Explainability methods for Vision Transformers (ViTs) at DIDSR, U.S. FDA (April 9, 2025).
Awards and Grants
- UM-IHC Travel Grant 2025 for ECAI '25
- US-RSE Travel Grant 2023
- STEMTrek ScienceSlam@SC21 Challenge Award
- High Achievers Award and Scholarship Grant (18th Position throughout Pakistan) at Secondary Level (2010)