Teaching machines to see and understand medical images — from brain MRI to agentic AI for clinical decision support.
Dr. Humera Tariq is an Associate Professor of Computer Science at the University of Karachi, working at the intersection of medical imaging, deep learning, and agentic AI. Her research teaches machines to segment and interpret clinical images — from brain MRI to quantitative bone-marrow MRI — and explores how AI agents can safely support diagnosis and clinical decision-making.
Her path bridges engineering and computer science: a B.E. in Electrical Engineering from NED University, followed by an MCS, MS/M.Phil, and PhD in Computer Science at the University of Karachi, with doctoral work on skull stripping and segmentation of brain MR volumes. She has held two international postdoctoral positions — biometrics and recognition at the University of Sassari in Italy, and MRI with deep learning at the University of Michigan in the USA.
Recent work spans deep-learning bone-marrow segmentation for myelofibrosis patients (published in Frontiers in Oncology and Magnetic Resonance in Medicine) and stateful agentic-AI workflows for stroke imaging and oncology real-world evidence. She is an NVIDIA DLI Certified Ambassador and a Meta AI Master Trainer for Pakistan, an ACM / ACM-W / SIGGRAPH member, and a reviewer for journals including NeuroImage and JMRI — regularly delivering Python and PyTorch deep-learning workshops from Karachi to EPFL in Switzerland.
Across the pipeline from raw medical images to safe, structured clinical understanding — and building the people and tools to get there.
Brain MR volumes, skull stripping, and quantitative fat-MRI bone-marrow segmentation using U-Net and Attention U-Net architectures.
Reproducible, end-to-end PyTorch and TensorFlow pipelines for clinical imaging, from data retrieval to validated segmentation outputs.
Stateful agent workflows — infarct core & penumbra segmentation on NCCT/DWI, and oncology real-world-evidence trial emulation.
Log-polar representations for biometric verification using contrastive learning and cosine-similarity face matching.
OpenGL/WebGL, 3D models and simulations, and voxel-processing pipelines for brain-volume measurement.
Prompt engineering and responsible-AI training as a Meta AI Master Trainer and NVIDIA DLI Ambassador.
Peer-reviewed work in medical imaging, deep learning, and applied machine learning.
A full publication list is available on GitHub and LinkedIn.
Invited talks, conference sessions, and hands-on deep-learning workshops delivered across Pakistan, the USA, Canada, and Europe.
Open to research collaborations, speaking invitations, workshops, and supervising students in medical imaging, deep learning, and AI.