Cirrus: Serverless Distributed Machine Learning
This project involves developing a serverless distributed machine learning framework to lower datacenter costs and achieve better utilization of resources. The idea is computers sometimes have unused computing resources that cannot be used up by conventional tasks, such as VMs, but these unused resources can be used by special serverless tasks, and we intend to write a framework for machine learning that takes advantage of serverless computing. This summer I will work on a python interface and a deep learning model existing Cirrus code.
Message To SponsorWorking with URAP so far has taught me the details of C++, as well as how to work efficiently in a group research environment. Through the Fall and Spring, I have worked with various existing ML frameworks, such as Tensorflow and Vowpal Wabbit, and learned how to write code and run experiments in an organized, documented fashion. These experiences are invaluable to me, and my work will hopefully help computer scientists down the line. I am excited and grateful to have been able to continue this work through the summer!
Major: Computer Science
Mentor: Randy Katz, Electrical Engineering and Computer Science
Sponsor: Chandra Research Fellows - Chandra Fund