Ali Anwar

CSENG Computer Science & Eng
College of Science & Engineering
Twin Cities
Project Title: 
Scalable and Resource Efficient Machine Learning

These researchers are using MSI for three projects:

  • Scalable Federated Learning: This project aims to pioneer new distributed machine learning methodologies scalable to millions of edge devices, such as cellphones and IoT sensors. A key focus is exploring the challenges posed by data and resource heterogeneity in federated learning environments. Utilizing MSI resources, the researcher have been able to simulate the operation of thousands of devices to conduct extensive federated learning experiments, which is vital for our research.
  • Model Storage: The goal for this project is to develop an efficient storage system specifically tailored for machine learning models. This involves analyzing a substantial volume of ML models to devise a system that can store ML data in a cost-effective and scalable manner. The computational and storage capabilities of MSI are crucial for handling and processing the large datasets involved in this project.
  • Energy and Resource-Efficient Inference for Large Language Models (LLMs): This project explores various system-level techniques to reduce the resources and energy required for large-scale LLM inference. The objective is to design and implement a pipeline that surpasses current practices in terms of efficiency. The processing power available through MSI is instrumental in experimenting with and refining these advanced models.

Project Investigators

Ammar Ahmed
Ali Anwar
Faaiq Bilal
Samuel Fountain
Connor Howe
Qi Le
Xinran Wang
Are you a member of this group? Log in to see more information.