EuroHPC Recognized Four Researchers for Supercomputing Work
The awards highlight strategic resource optimization in seismic research and multilingual language modeling.
Updated on Sept. 25, 2026 in Artificial Intelligence
Live Poll
Do you believe investments in large-scale supercomputing projects provide meaningful benefits to the public?
EuroHPC User Days 2026 honored four researchers in Dublin for their contributions to scientific computing and AI. The recognitions spanned both project-based resource optimization and academic research papers.
Why it matters
These awards spotlight the critical need for efficient supercomputing usage, where researcher proficiency in resource allocation directly dictates the scale of scientific output. As high-performance computing demand rises, these projects demonstrate how targeted hardware usage enables complex seismic mapping and advanced model training.
Award-winning projects demonstrated massive-scale compute utilization, with Stephen Beller managing 489,000 nodes for seismic research and Thomas Saillour consuming 160,000 MeluXina CPU node hours and 240,000 MareNostrum5 GPP node hours.
The players
Stephen Beller
A researcher recognized for managing 489,000 nodes in seismic data computational work.
Thomas Saillour
A computational expert honored for optimizing high-density node hours across multiple supercomputing architectures.
Carolina Oliveira Costa
A scientist focused on algorithmic solutions to catastrophic forgetting in multilingual AI models.
Dr. Rodrigo Bartolomeu
A researcher specializing in scaling P3M methods to improve force calculation efficiency on multi-GPU systems.
EuroHPC
The European High Performance Computing Joint Undertaking, which coordinates regional supercomputing infrastructure and policy.
The details
The recognized research targeted two primary bottlenecks: data-heavy seismic simulations and large-scale language model stability. Carolina Oliveira Costa developed methods to mitigate catastrophic forgetting—a phenomenon where neural networks lose previously learned information when trained on new data—within models supporting all 24 official EU languages. Meanwhile, Dr. Rodrigo Bartolomeu focused on scaling Particle-Particle Particle-Mesh (P3M) methods, an algorithm for calculating gravitational forces in n-body simulations, across multiple GPUs.
Timeline
Spring 2026: Researchers were invited to submit papers for consideration.
Week of Sept 25, 2026: EuroHPC User Days conference took place in Dublin.
The Tech Race
These awards track the success of the EuroHPC infrastructure program in fostering specialized user expertise on European hardware. This shift toward resource-aware engineering is a critical counter-balance to the growing computational requirements of large-scale AI and physics simulations.
These methodologies directly influence the training efficiency of large language models intended for regional accessibility in multiple languages. The demonstrated scaling techniques provide a roadmap for researchers looking to maximize limited compute budgets on shared supercomputing architectures.
The takeaway
The research highlights that computational efficiency is becoming as vital as raw power in modern AI development. Watch for future EuroHPC project reports to see how these scaling techniques influence the performance benchmarks of large multilingual model deployments.
Further reading
Explore the latest trends in high-performance computing and model efficiency in our Artificial Intelligence section.
Source note: This article includes information reported by HPCwire.
Live Poll
Do you believe investments in large-scale supercomputing projects provide meaningful benefits to the public?
