Computer Science, Special Report
October 2022

Efficient and inclusive

Argonne computer scientist works on power and processor efficiency while diversifying the workforce.

The Argonne National Laboratory-led Threadwork project is helping design high-energy physics detectors. Pictured here, the Compact Muon Solenoid (CMS) detector, an international collaboration that includes Argonne National Laboratory and Fermilab. Image courtesy of CMS partnership.

Editor’s note: Listen to the Valerie Taylor interview on Science in Parallel.

Argonne National Laboratory’s Valerie Taylor has built her career on improving the speed and efficiency of parallel calculations – fundamental advances that support large-scale computational science at the Department of Energy national laboratories.

In parallel to her computing achievements, she’s worked to lift historically marginalized groups in computing.

“Each community has unique experiences,” Taylor says. “But the commonality has to do with culture. Understanding culture requires intentionality. Not everyone has a Black child, Hispanic child, or a Native American child” in their lives and might not hear the day-to-day stories of how those people were discouraged from a career in math and science.

Valerie Taylor.

Taylor does much of her advocacy work through the Center for Minorities and People with Disabilities in IT, or CMD-IT (pronounced “command-it”), which she co-founded in 2011. She now serves as CEO and president. At the time, programs to improve gender diversity in computer science were gathering momentum.  Taylor and her colleagues wanted to build complementary opportunities to support computer scientists from underrepresented communities.

CMD-IT’s primary program is the annual Tapia Conference. The 2022 event, last month in Washington, drew 1,700 registrants and offered students and diverse professionals opportunities to connect with each other, obtain advice from diverse computing leaders, and hear inspiring talks.

CMD-IT has also launched a data-driven, National Science Foundation-funded project called the LEAP Alliance, aimed at broadening participation in computing. It focuses on increasing representation among research university computing faculty to improve diversity across computing. Using statistics from Jeff Huang at Brown University, the alliance has identified schools that produce the most computer science professors and launched an initiative to boost diversity among their graduate students. At institutions with higher student diversity, the project works to expose them to academic careers. At other institutions where undergraduates are most likely to eventually become faculty, the alliance is launching initiatives to retain students from underrepresented communities.

At Argonne, Taylor directs the mathematics and computer science division, managing research teams that work on numerical methods, optimization and artificial intelligence and deal with data management and storage, system software and advanced computer architectures. Much of the research is fundamental, but with important applications in science areas such as climate modeling, materials discovery and quantum information sciences.

‘Performance analysis allows you to combine the interest in hardware, software and the applications.’

Taylor is the principal investigator for Threadwork, among 10 DOE projects awarded a total of $54 million in 2021 for microelectronics co-design research. Her team of materials scientists, engineers, computer scientists and physicists from Argonne and Northwestern University has embarked on innovative ways to develop devices from new materials. The project ditches traditional approaches in which materials scientists might work most closely with chip engineers and software developers might talk with domain scientists – for a multilevel one, Taylor says. “It doesn’t have to go through your different levels of the system devices, the compute system, the software stack and the application stack. So we have applications, staff and researchers talking with materials researchers. It’s not just each adjacent level talking to each other.”

A key application: Designing detectors for high-energy physics. As lasers become more powerful, they will require faster detectors that can capture data and connect to computers for in situ processing – as the calculations progress – and further analysis. New materials for novel devices such as memristors could help researchers build terahertz interconnects, which can bridge optical and electrical signals.

Before joining Argonne in 2017, Taylor spent more than 25 years at Northwestern and Texas A&M universities. She began collaborating with Argonne soon after joining Northwestern, working on performance analysis and efficiency in scientific computing systems. “Performance analysis allows you to combine the interest in hardware, software and the applications,” she says.

That lens has let her work on integrating processors into high-performance computers, including graphics processing units (GPUs), field-programmable gate arrays (FPGAs, flexible chips that can be programmed after they are manufactured) and new artificial intelligence accelerators. “You’re not just looking at performance of a CPU,” she notes. “You’re looking at the performance across the system and taking into account the complexity of what a node looks like. How are the nodes configured?” And with power efficiency, a researcher can look beyond individual chips or systems. “What about the impact on the environment? What’s the carbon emission?”

With machine learning and artificial intelligence, Taylor’s research interests and service passions have come full circle. She recently coedited “Transforming Trajectories for Women of Color in Tech,” a report from a committee of the National Academies of Sciences, Engineering and Medicine, after women in those fields met to discuss and address career concerns.

Most tech-career research lumps women into one group. STEM fields need studies on the “nuanced and specific needs of different subgroups of women of color,” the report concludes. For example, Taylor says, did a program aimed at all women also support women of color? “Even among Blacks, did it work well for Black men? What about Black women?” It’s not about blindly applying AI to racial injustice. “But we have a lot of data that we can leverage to look at these trends, to better understand what’s going on with different communities.”