My research interests are in vision-language models, self-supervised learning, and 3D perception. I work on adapting vision-language models without target annotations, learning grounded representations from child-view video, and reliable visual recognition under domain shift.
I received my B.Sc. (Hons) in Computer Engineering from the University of Peradeniya, Sri Lanka.
MedCache: Training-Free Spatially Aware Caching for Accelerated Medical Video Generation
Ufaq Khan, Umair Nawaz, Sathira Silva, Numan Saeed, Muhammad Umer Sheikh, Muhammad Bilal, Junaid Qadir, Lena Maier-Hein, Yutong Xie, Muhammad Haris Khan
I am investigating how visual recognition systems behave when field conditions obscure diagnostic evidence or unfamiliar symptoms appear. This work focuses on cross-source recognition, calibration, and selective prediction, alongside the development of MaizeShift-RBench.
I contributed to the technical research plan and methodology of the successful SafeMaize proposal, funded by the Gates Foundation through IAAI. I also contribute to field-data collection planning, agronomic annotation, and expert validation with project partners.
Experience
May 2025 – Present
Research Engineer · MBZUAI
Computer Vision Department · Dr. Muhammad Haris Khan
Vision-language adaptation, child-view video learning, and reliable visual recognition.
Jun 2024 – Apr 2025
Research Assistant · MBZUAI
Computer Vision Department · Dr. Muhammad Haris Khan
Unsupervised CLIP adaptation with visual and textual prototypes, fine-grained interactions, and self-training.
Dec 2022 – Mar 2023
Computer Vision Research Engineering Intern
Vega Innovations · Autonomous Vehicle R&D
Transformer-based perception and real-time vision on NVIDIA DRIVE PX2 and Jetson TX2.
May 2021 – Sep 2021
Teaching Assistant · University of Peradeniya
Programming Methodology (CO222). C programming labs, quizzes, and individual tutoring.
