Science Node recently talked with Sharon Broude Geva, director of Advanced Research Computing (ARC) at the University of Michigan (U-M) about the challenges of enabling data-intensive and computational research at a highly-interdisciplinary Research 1 university.
Geva sees an inclusive future that brings the benefits of advanced computing to more people, including women and other underrepresented groups, and to more disciplines than the traditional engineering and physical sciences of the past.
Let’s start at the beginning--how did you get involved in advanced computing?
I have a Ph.D. in computational chemistry. Even before that, I was heavily involved in HPC, which was then called supercomputing. I started off building supercomputers, writing compilers for them, writing applications for them. After my post-doc, I spent almost a decade in industry where I led software project and design teams.
I specialized in usability, which I really liked. It brought in social science, people, and technical aspects and tied into lots of things I had seen while working on supercomputers. When I came to the University of Michigan (U-M) , they liked the idea of having someone who could talk to faculty and understand their computational needs, while also running a research project to find out what was missing from the university’s research computation environment.
At the time, there were scattered clusters and expertise across U-M for specific groups, but there was no central organization that could serve everyone. With more than 61,000 students and 6,300 faculty members across three campuses and representing a variety of disciplines, it was clear we needed to have a better use of resources that would spread the capability across everyone at the university.
What does advanced computing for an entire university look like?
The University of Michigan is very interdisciplinary and our computing resources have to reflect that in every way. So our question was, “How do you make it so researchers have a place where they can think things out together and learn from each other?”
In 2014, we created this new, consolidated organization that included a very large emphasis on all aspects of advanced research computing. We wanted to bring in a whole catalytic and support ecosystem with a focus on innovation and multidisciplinary collaboration.
The ecosystem includes academic seminars, symposia, and working groups for computational and data-intensive researchers and students, student clubs and fellowships, catalyst grants for seed projects, and much more.
Everyone up to the associate vice president I report to partners with units to identify ways in which we can help. Our consulting and training unit schedules 3,000 face-to-face appointments a year. Any researcher can ask for an hour a week, for help getting over hurdles, learning how to analyze, manage, and clean data, or even to optimize code for HPC.
We offer a vast breadth of workshops, everything from basic stats regression all the way through to neural networks and machine learning and specific software such as TensorFlow. We also include more technical training like working with the Linux line command, how to get on a cluster, things like that. We try to make sure people have the knowledge they need.
You’re a big supporter of the Women in HPC (WHPC) group. How did that come about?
I first joined WHPC because I’m used to being the only woman in a room.
One of the things WHPC really works on is making it clear that not only women should be part of WHPC. We have to stop thinking that women are responsible for dealing with the fact that there aren’t women in the field or that women aren’t retained. Getting male allies and male advocates is very important.
I’m really excited about the WHPC mentoring program that’s just starting up—I’ve signed up to be both a mentor and a mentee. Through my budget at U-M, I support WHPC and Girls Who Code. I’m on quite a few steering committees nationally and internationally, and I’m very careful to point out when there is not enough speaker or committee diversity. As soon as I mention it, I’m usually asked to provide suggestions, and I can then point to very qualified women that the program committee may not have thought about.
I try to live my life in a way that is a role model as much as possible. I try to learn from every role model I see. I try to stand up for diversity in every way. It’s not just about women. Diversity is about the inclusion of any kind of social, gender, or ethnic diversity, different backgrounds, abilities and different educational backgrounds, to name but a few.
Every time a younger woman comes up to me and says, “Women in HPC completely changed the way I think about things,” I realize I am not alone in this community. This is a bigger set of issues and there’s a bigger group of people working on these issues. It’s truly inspiring. My reason for joining has grown even stronger over the years.
U-M has one of the first local chapters of WHPC. What’s the benefit of that?
The chapters bring a local participation that many women or men can’t get from the big national and international meetings. Not everyone goes to the SC conferences.
Our chapter at U-M is meant for the academic more than the IT side. It brings together students or post-docs with faculty members who are women and offers the possibility of getting to know them and see them as role models. When you’re in a room full of women, it is easier to talk about the things that you do, including your strategies for dealing with challenges in your work, your work-life balance strategies, the experience and capabilities you bring, and your career aspirations.
The chapters’ program is in its pilot phase globally and will be opening for everyone soon. We wanted to test a small set of chapters and affiliates so we could learn what needed to change in the program before we expanded. But we’re also telling people, “Don’t wait. Start thinking about how it will work on your campus or organization.”
From your birds-eye view of the computationally enabled research that’s taking place at U-M, what trends do you see emerging?
The intense users of HPC used to be from chemistry, astrophysics, nuclear engineering, and a few more domains. But now it’s no longer just a handful of research areas. People who never thought they’d be doing advanced research computing, now this is one of the tools they have.
But people coming in these days, even if they’re from one of the traditional fields, don’t necessarily have the coding background that people did twenty years ago. They want to spend their time on research, not coding. That makes the consulting and training aspect of HPC even more important. We have to help people learn how to do things and make the tools a lot more usable.