Yusuf Ali

Hi! I am a PhD student at the School of Interactive Computing, Georgia Tech working with Prof. Zsolt Kira . My research focuses on improving foundation models at test time without relying on large-scale data collection, heavy fine-tuning, or sacrificing general model capabilities in diverse application domains. Most recently, I have explored this problem in the context of embodied VLA models. I have also studied the capabilities and limitations of frontier multimodal LLMs in settings that require long-horizon memory and robust performance in low-data domains. My current work investigates how minimal and targeted test-time updates, particularly through reinforcement learning, can improve VLA performance and multimodal video reasoning. Please reach out if you would like to discuss ideas or collaborate !