Supercomputing Academy Offers Path to HPC Expertise

Keyvisual image main
HLRS's training programs helped Alaa Bejaoui und Maha Badri to develop essential skills for parallel programming and artificial intelligence that they will continue to use in their doctoral studies. Photo: Yüksel Patir

Two recently certified HPC Experts reflect on the benefits of HLRS’s training programs for their education, research, and career development.

Maha Badri and Alaa Bejaoui were both born in Tunisia, but they also have much more in common. As young engineers who in 2023 completed their master’s studies in aerospace engineering at the University of Stuttgart, they share a passion for mathematics, artificial intelligence, parallel programming, and high-performance computing. During their bachelor’s studies, Badri and Bejaoui began to take advantage of the comprehensive high-performance computing (HPC) training program offered at the High-Performance Computing Center Stuttgart (HLRS). They became so intent on gaining a thorough understanding of how to program HPC systems that during their master’s studies they became the first participants in HLRS’s Supercomputing Academy to be awarded the certification “HPC Expert.”

From the beginning of their university studies, the engineers recognized the importance of parallel programming, but saw that the necessary training would not take place within their academic program. “When we began to study in 2013, we quickly noticed that engineers often did not place a high value on programming,” Bejaoui recalls. “Because of this gap we decided to take as many workshops at HLRS as we could.” Initially, they took courses in HLRS’s core HPC training program. After spending time in Canada to study artificial intelligence at the Polytechnique Montréal, they returned to Stuttgart in 2020 and discovered HLRS’s newly founded Supercomputing Academy, and welcomed the opportunity to dive deeper. According to Badri, “We already had basic knowledge, but we wanted to expand on it.”

A deep dive into HPC and AI

Whereas the HPC classes they had taken previously at HLRS were held over 3-5 days, courses offered by the Supercomputing Academy typically extend over approximately 6 weeks in a blended learning format, enabling participants to learn at home in parallel with their other professional responsibilities. It is possible to take individual courses, but for participants interested in a comprehensive training experience, the Supercomputing Academy offers certification as an “HPC Expert.” The program’s content is modular, enabling participants to select from a menu of courses to explore their individual needs and interests. Badri and Bejaoui elected to pursue certification in the category “HPC Developer.” In the process, they gained knowledge about MPI, Open MP, communication, and node-level performance. Because of their interest in artificial intelligence, they also took courses in data management and data analysis.

“From the beginning our goal has been to complete doctorate degrees,” Badri explained, “and the idea (to pursue Supercomputing Academy certification) was to prepare ourselves as well as possible… The fact that the program requires tests was important for us, because taking a class without a test is not the same as when you study to pass a test. You go much deeper and learn more intensively.”

Because Badri and Bejaoui were active in the Supercomputing Academy during the COVID-19 pandemic, they unfortunately missed the personal contacts with instructors and students that they enjoyed during earlier on-site courses at HLRS. Nevertheless, the design of the Supercomputing Academy was an ideal way for them to develop their HPC skills while also completing their academic program. Badri says, “I could concentrate on my master’s thesis during the afternoon and at night I would watch the (Supercomputing Academy) course videos… If everything had taken place in person, I can’t imagine how it would have worked.”

A strong foundation for career development

Now that they have completed the Supercomputing Academy’s training program, Badri and Bejaoui recognize that the experience has had many benefits. “It was only through the Supercomputing Academy that we gained a deep understanding of HPC hardware, and this has made us more deliberate in our programming, particularly with respect to parallelization,” Badri says. In practice, they used their new skills to apply MPI for Python programming, making it possible to parallelize an approach that combines simulation and reinforcement learning.

During job interviews, they also sensed that the certification offered through the Supercomputing Academy made them more attractive job candidates. Badri recalls, “In all of my interviews, it was a feature in my resume that impressed potential employers. They always asked what kind of certificate it was. Was it for a single course, or for a week-long course? When I explained all that we had learned, they were very impressed at how comprehensive the program was.”

Through the experience they have gained through their academic studies, job searches, and work at a startup company, Badri and Bejaoui sense that there is a growing demand for the training that the Supercomputing Academy offers. “In fields such as aerospace engineering that rely on simulation, HPC is well known, as simulations are not possible without an understanding of HPC. With the rise of AI, however, other working groups are realizing that they need expertise,” Badri explains. Bejaoui agrees, and thinks that the training they received has positioned them well. “HLRS and the Supercomputing Academy are good partners for developing these skills,” he says.

The two are currently looking forward to starting their doctoral studies, which will take them in exciting new directions. Badri will soon join the Potsdam Institute for Climate Impact Research, where she will use artificial intelligence to develop vegetation models that consider interactions between the plant kingdom and the atmosphere. Bejaoui has secured a doctoral position at the Charité University Hospital in Berlin, where he will collaborate with medical doctors to develop methods that use artificial intelligence to interpret CT scans, tools that could lead to better patient care. Both anticipate that the HPC expertise they have gained will continue to be necessary in their respective projects.

Over the longer term, Badri and Bejaoui also hope to carry their knowledge back to their home country, Tunisia. “We have good scientists, very smart minds … with expertise in mathematics and physics. And now computer science and HPC are coming. I feel like we now have the complete chain, from mathematical modeling, to implementation, to optimizing the implementation on hardware,” Bejaoui says. “We want to build bridges between Germany and Tunisia, and Germany could also benefit.”

New HLRS training concept integrates Supercomputing Academy and compact courses

HLRS continues to develop its training concept to offer course participants a menu of options that best address HPC-users’ professional interests and needs. According to Lorenzo Zanon, head of the HLRS Department of Training Scalable Algorithms, the center has plans to combine successful elements of the Supercomputing Academy and its compact HPC training courses: "This could make it possible, for example, to complete exams at the end of compact training courses that will count toward the HPC Expert certification. Conversely, HPC users could register for blended learning courses in the Supercomputing Academy without the requirement of taking a final exam." In this way, HLRS aims to provide a flexible training program that is as accessible as possible. More news about this concept should be available in 2024.

Learn more about HLRS’s training program at www.hlrs.de/training. Additional information about the Supercomputing Academy is available at www.supercomputing-akademie.de.

Podcast

Click here to listen to a conversation with Maha Badri and Alaa Bejaoui in a podcast from SICOS-BW. (The interview was conducted in German.)

Christopher Williams (interview by Andreas Wierse)