The Data Systems Lab focuses on developing software systems and libraries that empower developers, data scientists, and machine learning experts to effectively utilize expanding datasets with exceptional performance. Through user-friendly interfaces, our systems eliminate the need for users to grapple with the intricacies of scaling to distributed systems, multi-core servers, massively parallel GPU systems, and the cloud. Our research efforts are centered around building efficient systems for machine learning model training and inference, big data processing, data management, and storage.
You can find more about the DREAM lab at https://dream.cs.umass.edu
Project Description:
Our group is developing systems for efficient training of Graph Neural Networks. The students will run these systems, measure their performance, and develop some of their components.
Learning Objectives:
Students will become familiar with the tools use to train deep neural networks. They will learn how to run machine learning pipelines and measure their performance. They will be exposed to the inner workings of large-scale multi-GPU clusters.
Skills needed:
Be able to use basic linux command-line tools to organize and process data and to install software.
Be ready to learn how to develop software systems at a scale that goes way beyond the typical undergrad assignment. Be ready to learn some basic concepts of deep neural networks.