CSE
CSE
CSE CSE

For Black History Month, CSE Spotlights Faculty and Alumni in Academia

  Bookmark and Share

Marcus Darden | Odest Chadwicke Jenkins | Jason Mars
Kyla McMullen | James Mickens | Kunle Olukotun

Jason Mars

Jason Mars is an Assistant Professor at CSE, where his research interests include cross-layer systems architectures in both software and hardware, data center and warehouse scale computer architecture, and hardware/software co-design focused on native application performance, energy efficiency, and system utilization, particularly in the context of the latest innovations in microarchitectural design, runtime systems, and cloud computing.


Jason Mars

Prof. Mars' work is noteworthy because it takes the path of cutting across many of the traditional areas of specialization in computer science, such as system architecture, artificial intelligence, or operating systems and compilers, amongst others. Prof. Mars looks at how these realms interact in the overall context of creating intelligent and efficient cloud computing platforms that will be responsive to users' needs.

"We have to discover ways to re-think the system architecture of data centers to be able to provide the necessary computing efficiently enough to service the millions and potentially billions of users who are interacting with cloud services in increasingly sophisticated ways," says Prof. Mars.

This will require, he says, a new layer of abstraction that lives between the applications that run on cloud systems and the underlaying hardware. Since joining CSE in 2013, he and his collaborators have launched a number of projects in this area, one of which has grown into a commercial venture.

video
Prof. Mars and his collaborators on Protean Code

In 2014, Prof. Mars and his collaborators announced Protean Code, a new technique for recompiling the applications that run on data center servers and introducing that updated code as needed. Protean Code allows data centers to dynamically rebalance the use of system resources as needs dictate, and it is unique in doing so without a performance penalty during recompilation.

In 2015, after recognizing that unique processing requirements existed for processing media-intensive user interactions, Mars and others introduced an open-source intelligent personal assistant (IPA) platform that included speech recognition, image matching, natural language processing, and a Q&A system.

Because commercially available IPAs, such as Apple's Siri and Microsoft's Cortana, were essentially "black boxes" that could not be manipulated, the Michigan researchers' IPA system was intended to be used primarily by those who needed to understand the inner workings of such systems and how they would interact with data center resources.

video
Prof. Mars and his collaborators on their Intelligent Personal Assistant software

Prof. Mars and others also recognized that cloud-based IPA systems represented an enabling technology in the next generation of smart applications, and as a result, Prof. Mars engaged in two additional notable projects.

One is a partnership between Michigan and IBM, led by Prof. Satinder Singh Baveja, aimed at creating an advanced conversational computing system. Announced in early 2016 with $4.5 million in funding from IBM, the project is envisioned to create a system that will engage in natural, multi-turn conversations with users, similar to human-to-human conversations. It will ultimately be deployed as a "digital advisor" for computer science students, advising on routine matters and handing off conversations to human advisors for more specific and personalized counselling.

In another ongoing project, Prof. Mars, along with his key collaborator Prof. Lingjia Tang and some of their current and former students, commercialized their IPA technology through the startup Clinc in the latter half of 2015. In addition to working with partners who will use their technology, Clinc has introduced Finie, an intelligent personal financial assistant that helps users to talk to their bank accounts in a natural and conversational way to get real-time and instant financial insights.

Prof. Mars received his PhD in Computer Science at The University of Virginia in 2012 and joined the faculty at Michigan in 2013. From 2012–2013, he was the Peggy and Peter Preuss Faculty Scholar in the CSE department at the University of California, San Diego. He has served as a visiting scientist at Google, which involved investigating opportunities to improve efficiency of Google’s backend infrastructure.

Prof. Mars work has been spotlighted as an IEEE Micro "Top Pick" in 2011, 2012, and in 2016. He received an NSF CAREER Award in 2016. He served as Program Chair for the 2015 International Symposium on Code Generation and Optimization (CGO).


Posted: February 3, 2017