Soroush Mahdi

I am Soroush Mahdi, a researcher in Artificial Intelligence with a focus on 3D computer vision, vision-language models, and trustworthy deep learning.

I earned my M.Sc. in Artificial Intelligence from Amirkabir University of Technology (Tehran Polytechnic), where my thesis focused on improving the robustness of deep neural networks against adversarial attacks, under the supervision of Prof. Maryam Amir Mazlaghani.

Currently, I am a research assistant at the Autonomous & Intelligent Systems Lab (AISL), where I focus on 3D vision and deep structure-from-motion (SfM) models. My recent work, Evict3R, proposes a training-free token eviction approach for efficient large-scale streaming visual geometry transformers.

Research Interests

• 3D Computer Vision
• Vision-Language Models (VLMs)
• Computational Neuroscience
• Trustworthy & Robust Deep Learning
• Generative AI

I am particularly interested in the intersection of 3D vision, brain-inspired modeling, and generative deep learning. I aim to pursue a Ph.D. in Deep Learning and Computer Vision to continue advancing research in these areas.

If you are interested in research collaboration or academic discussion, please feel free to get in touch.