This research project seeks to innovate at the intersection of advanced vision transformer architectures and dynamic data processing. It aims to develop a versatile framework capable of adapting to various real-time applications, addressing the need for balanced support across different research branches and paving the way for breakthroughs in efficient data processing and analysis. Unlike existing works that focus on either improving the transformer architectures for specific domains or developing efficient and effective techniques for dynamic data processing, this research project aims to integrate these two aspects and develop a learning-based framework that can adapt to various real-time applications. The concept of Age of Information (AoI), which has gained significant attentions in networked systems, is incorporated to improve the timelessness of the information flow of the transformer architecture as well as to support real-time adaptive update architecture parameters, which combined address the challenges associated with the computational overhead of vision transformers in general.