Ayesh Abu Lehyeh

I am a Ph.D. researcher in Computer Science at the University of Vermont, specializing in cross-view image geolocalization. My research focuses on visual geolocalization from ground-level imagery using overhead and map-based representations, with emphasis on achieving real-time, fine-grained localization at sub-meter accuracy in real-world settings. My work advances learning-based localization models that extend beyond retrieval through continuous refinement via cross-view image matching, geometric reasoning, and uncertainty-aware modeling. These approaches enable robust localization under significant viewpoint changes and environmental variability. I also investigate multi-modal representation learning and vision–language models for geolocalization, aiming to enable semantic reasoning about location and provide interpretable explanations beyond black-box predictions. Recently, I have been exploring generative modeling approaches, particularly flow matching, as a principled framework for modeling spatial transformations in visual geolocalization.

In parallel with my academic research, I have worked on applied machine learning problems in collaboration with industry, including wireless sensor network localization and data-driven modeling for semiconductor manufacturing at GlobalFoundries.

I am interested in interdisciplinary collaborations that address practical challenges in visual geolocalization and spatial reasoning, and I am always open to discussing research ideas or potential collaborations.