Md A Rahman

PhD student in Computer Science at the University of Texas Rio Grande Valley. I work on multimodal AI systems that can be trusted with real decisions — vision-language models, tool-using agents, and retrieval — focused on grounding, calibration, and abstention: making a system cite its evidence, quantify its uncertainty, and refuse the questions it cannot support. I test these ideas where the ground truth is physical and a wrong answer has a cost: satellite and aerial imagery, LiDAR, UAV sensing, and large engineering document collections. Interests: vision-language models, agents and tool use, retrieval-augmented generation, uncertainty quantification, hallucination and abstention, benchmark design, and geospatial machine learning.