Internship Program Fosters HPC Talent
One of the most pressing concerns in HPC circles continues to be a lack of qualified entrants to the field. Not to say there isn’t talent — the popularity and success of the student cluster challenges attest to that — but it’s not enough to span the gap. That’s why efforts to facilitate this unique skill set are so crucial. One initiative seeking to do just that is the Texas Advanced Computing Center’s (TACC’s) STAR Scholars Internship Program. The program, which seeks to bolster a computationally literate workforce in HPC and scientific visualization, just celebrated its one year mark.
The idea for this joint academic-industry program — which grew out of TACC’s industrial affiliates program, the Science and Technology Affiliates for Research (STAR) — was to identify UT Austin students with a background in computational science and connect them with a TACC mentor to work on a project of interest to industry. The industry partners, for example BP and Chevron, provided financial backing, while TACC taught the students the HPC skills necessary for today’s job market.
It was actually Keith Gray, an HPC manager at BP, who provided the spark for the project when he approached TACC to sponsor and train three UT Austin students in advance of them signing onto internships with the oil and gas company.
“Many of the skills required to succeed at BP are not taught in the classroom — that experience comes from hands-on work,” Gray stated. “By sponsoring interns, providing them a TACC mentor and broad set of problems to work on, we’re ensuring that they are able to solve business-scale problems.”
Madhav Gupta, a junior studying Electrical and Computer Engineering, was one of four students selected for the program in its first year. Gupta expressed enthusiasm for the chance to collaborate with his professor, Ritu Arora, on a real industry project, and for the career-boosting potential as well.
“This internship gives me programming experience I wouldn’t normally get in my academic classes,” he said.
There’s also a very real possibility for graduates of the program to be hired onto these companies.
“Corporations have many open positions requiring HPC expertise,” TACC’s Industrial Programs Director Melyssa Fratkin, said. “TACC is working with industry to fill this gap, giving them a competitive edge for the 21st century economy.”
In the inaugural year of the program, the four students who were selected worked 10-20 hours a week on a research project during the academic year and were paid a small stipend.
Given the nature of the program, it packs a lot of learning and hands-on experience into one year. After a winter-break long crash course on basics of HPC, Gupta jumped into a pretty tricky problem area, creating a tool to hide some of the complexity of HPC coding. The goal was to develop a computing tool that converts existing C or C++ code into a parallel variant (MPI, OpenMP and CUDA) that is suitable for use on supercomputers, like TACC’s Stampede system.
“When scientists use TACC’s resources, they are more interested in getting their science done, rather than getting bogged down in the low-level details of programming,” Arora explained. “The tool that Madhav and I are working on acts as a bridging tool, so that researchers can run their code easily on our supercomputers.”
According to Gupta and Arora, work on the tool is progressing nicely, so they plan to continue their collaboration through the next school year.
Other first year applications include algorithms that predict mutations in cancer genomes; a portal for the Sustainable Places Project, which helps city planners understand the impact of various growth patterns such as increased traffic and pollution; and using benchmarking to boost the performance of TACC clusters.