Special Report
January 2015

Careers that compute

DOE computing jobs go begging for applicants, study finds.

Programs like the Department of Energy Computational Science Graduate Fellowship (DOE CSGF) provide essential training to young investigators like Sam Skillman, a DOE CSGF alumnus and postdoctoral researcher with the Kavli Institute for Particle Astrophysics and Cosmology. In one day last year on the Titan supercomputer, Skillman and colleagues simulated the 13.7 billion-year-old expanding universe, including this small subset of the particles showing a massive galaxy cluster at the image’s center. Particles are colored by their incoming (blue) or outgoing (red) velocity with respect to the cluster.

The demand for advanced scientific computing experts far outstrips the supply of well-trained computer scientists to fill those needs. To address this workforce gap, the DOE’s Office of Science asked the Advanced Scientific Computing Advisory Committee (ASCAC) last year to identify the potential obstacles to filling current and future professional computer and computational science positions and to recommend solutions.

At the recent ASCAC meeting at SC14 in New Orleans, subcommittee Chairperson Barbara Chapman of the University of Houston reported the group’s findings. After collecting data from DOE’s national laboratories, university computing programs and previous studies of workforce preparedness, the subcommittee identified several trends that contribute to difficulty attracting and retaining computer science professionals.

The group found that many of the factors contributing to difficulty filling such positions are not unique to government science hiring. For instance, relatively few U.S. citizens obtain degrees in the computing sciences. As a result, many of the national labs report hiring more and more foreign nationals, which account for more than half of the graduate students in Ph.D.-granting computer science programs.

This trend exacerbates the reported long lag between the time a position is posted and when it can be filled. The DOE Office of Science labs reported that it typically takes 100 days to fill a job, compared with 48 to 50 days to fill a similar position in industry. The lag is even more pronounced at DOE National Nuclear Security Administration (NNSA) labs, where jobs require U.S. citizenship. There, jobs can take 200 or more days to fill.

Two intersecting factors – high demand and low unemployment – make computing a hot field. Yet those who study a computational field tend to choose specialties such as artificial intelligence and robotics. There are fewer students choosing to train in areas such as algorithms, applied mathematics, data analysis and visualization, and high-performance computing systems, which are all in demand at the national labs.

‘This is an issue for the whole supercomputing community.’

Chapman points out that these jobs require experts cross-trained not only in the relevant computational sciences but also applied mathematics, statistics or traditional scientific fields such as chemistry and physics. Yet the subcommittee found that very few academic programs provide this kind of cross-disciplinary training.

“New graduates are generally not prepared to take on the demands of a multidisciplinary workforce,” Chapman says.

Further, women and minority groups are nearly absent from the ranks of computing professionals. The subcommittee found women comprise a low and declining percentage of computing graduates, with 17.2 percent of computer science and 18 percent of all computing doctorates. Less than 2 percent of computational science doctorates are awarded to Hispanic or African-American students.

“This is an issue for the whole supercomputing community,” Chapman says. “Meeting the mission-critical workforce needs of the national laboratories will require leadership to address this lack of diversity and to design outreach programs to attract a more diverse student population.”

Projections forecast an increase in the computer science workforce gap and persistent underrepresentation of women and minorities unless the nation takes action. Attracting and retaining excellent students of both genders and all ethnic groups is a challenge faced by many fields and is certainly not limited to computing, Chapman says.

The subcommittee suggested that DOE contribute in this area by increasing outreach to universities serving minority populations. They cited DOE laboratories’ successful engagement in graduate education through the Nuclear Science and Security Consortium, which includes seven major universities, five minority-serving universities and four DOE labs. This five-year program, designed to train a generation of nuclear scientists, has supported more than 100 students to date.

Another successful program, the DOE Computational Science Graduate Fellowship (DOE CSGF), should be maintained and even expanded, Chapman says. The DOE CSGF has been rated highly effective in multiple reviews, she notes, and by design is uniquely structured to train students in interdisciplinary knowledge and provide DOE lab experience, often leading to jobs and careers there. Unfortunately, the DOE CSGF doesn’t produce enough graduates to fully satisfy workforce demands. Therefore, the subcommittee recommends using the DOE CSGF program as a model that could be replicated to address specific interdisciplinary needs such as exascale algorithms and extreme computing.

The group also noted that many DOE laboratories have developed their own measures to help recruit and retain computational scientists. More could be done at the national level to “amplify and extend the effectiveness of their locally developed programs,” the subcommittee reported to ASCAC. The report suggested laboratories share information about those efforts.

The subcommittee also recommended developing a national program to increase DOE’s visibility on university and college campuses. They suggested that the advent of online education, massive open online courses and distance learning offers an opportunity for laboratories to collaborate with universities to teach in-depth classes in topics of interest to the labs and in specialties where their scientists offer world-class expertise.

Finally, the subcommittee suggested that DOE make laboratory careers more attractive. For instance, Chapman says, the DOE should examine the laboratory funding model and its impact on recruiting and retention, and also write uniform policies about sabbatical leave, family-friendly flexible workplaces and professional development opportunities.

Although government computing jobs will likely never be able to compete with the compensation available in industry, Chapman says, the labs do offer access to cutting-edge technology and interesting projects that can make positions more attractive.

“One thing we heard over and over again,” she says, “was that the key to recruitment is to get students into the labs so they can learn what exciting science is being done there, and that gets them hooked.”