AI Acceleration

Consider the challenge of deploying complex machine learning models on resource-constrained devices such as smartphones or sensors. These devices often struggle to meet the computational and memory demands of large-scale models. This issue is particularly pronounced in scenarios involving privacy-sensitive data or applications requiring rapid response times. The disparity between the computational capabilities of edge devices and the requirements of modern machine learning models represents a significant barrier to the widespread adoption of intelligent applications.

Large Model Distribution

Distributed training and inference of machine learning models present their own set of challenges. When attempting to distribute a large model across multiple devices, issues such as communication overhead and data heterogeneity can significantly impact performance. These factors can transform what should be a cohesive multi-device operation into a disjointed and inefficient process. The resulting inefficiencies can lead to suboptimal performance and increased operational costs, complicating the deployment of these models in real-world applications.

Our research group is committed to addressing these challenges through the development of innovative system solutions aimed at enhancing the efficiency and accessibility of machine learning technologies. We are exploring novel approaches to optimize machine learning models for practical applications, with a focus on mitigating computational and communication bottlenecks. Our objective is to extend the capabilities of advanced machine learning models to a broader range of devices and environments, thereby making intelligent decision-making more accessible and efficient across various platforms and applications.

Ph.D Students

Master’s Students