Muhammad Zawish

From 21/08/2023 to 30/11/2023

Open Ideas

Complexity-aware Collaborative AI over edge and cloud

Saint Louis University

I’m a PhD Candidate at the Walton Institute, South East Technological University, Waterford, Ireland. My research is focused on improving the energy and computational efficiency of AI models in the context of resource-constrained IoT-enabled edge systems. In particular, I investigate the research question on how can we achieve energy efficiency in contemporary AI models without sacrificing their value (accuracy). So far, my research have been published in high-quality peer-reviewed IEEE venues, such as IEEE IoT Journal and Magazine and IEEE Globecom Conference. I’m a student member of IEEE Communication Society and also serve as a volunteer/reviewer in top IEEE/ACM venues like IEEE WCM, IEEE TGCN, IEEE TII,  ACM e-Energy, ACM COMPASS, AAAI, IEEE Globecom, etc.

This project aims to develop an intelligent lightweight computing solution for deploying AI models, specifically deep neural networks, in complex computer vision applications on the Internet of Things (IoT). The purpose is to address resource constraints at the edge by delegating minimal computations to the edge, while ensuring ultra-low latency for time-sensitive applications. The challenge lies in avoiding the static and cumbersome re-configuration of AI models across edge and cloud when resources change dynamically. Unlike existing approaches, our solution will be intelligent, proactive, and dynamically aware of changing environment, such as varying network bandwidth in different locations or varying quality of input data. This will enable seamless deployment and optimal adaption of AI models at runtime, without the need for iterative re-profiling, ensuring efficient and uninterrupted service quality.

In the first half of fellowship, the following activities were performed:

• Engaged in extensive discussions with my host professor at Saint Louis University (SLU) during the initial fellowship period
• Focused on conducting a comprehensive literature review on various lightweight computing approaches and challenges that surpass the state of the art
• Explored the identification of gaps in the domain and strategized the development of a methodology for evaluating our approach
• Implemented and evaluated initial experiments towards energy efficient AI on resource-constrained edge
• Actively participated in networking activities within the Networks Research group at SLU to enhance collaboration and knowledge exchange
• Showcased my research by presenting at the IEEE ComSoc School Series, resulting in the honor of receiving the Best Presentation Award

• Development/Advancement of Innovative Technologies: Introducing new approaches to create lightweight and energy-efficient deep learning models
• Testing Technologies (Demo, Pilot…): Assessing the energy efficiency of deep neural networks before and after optimization on common edge computing hardware like Nvidia Jetson Nano
• Sound Scientific Validation: Ensuring the open access policy scientific publications, all publications will be made available in a timely manner via openly accessible repositories
• Strengthening Research Collaboration with the US/Canada: Establishing a long-term collaboration with PhD students in the Network Research Group at Saint Louis University
• Building Solid Connections and Partnerships in Europe and in the US/Canada: Collaborating closely with SFI VistaMilk Research Center in Ireland and the NRG group at SLU to establish strong connections and partnerships
• Fundraising (Proposals to Public Organizations) – Indicate, in the US/Canada or Europe: Successfully securing a travel grant to attend the IEEE ComSoc School Series at GeorgiaTech Research Institute
• Paper Submission for Further Publication – Indicate, Only EU Author(s), or Jointly with the Host Organization: Planning to submit a conference paper jointly with the host organization for further publication
• Conference Attendance with Paper/Poster/Proceedings: Participated in the IEEE ComSoc School Series on Next Generation Cloud Computing at GeorgiaTech Research Institute and presenting research findings