The Czech Institute of Informatics, Robotics and Cybernetics of the Czech Technical University in Prague (CIIRC CTU), in cooperation with the International Neurodegenerative Disorders Research Center (INDRC), has completed the development of new computing infrastructure for the CLARA Centre of Excellence. The new GPU cluster, with computing performance of up to 260 PFLOPS, represents the largest institutional installation of specialised computing capacity for artificial intelligence (AI) applications in the Czech Republic over the past year. It is now fully operational and is currently being used primarily in two areas. The first is the development of AI models that help researchers understand the molecular mechanisms underlying neurodegenerative diseases and identify compounds that could influence them. The second is the development of AI methods for robotics, with a focus on assistive technologies for people affected by neurodegenerative disorders.
The installation was completed in the second half of 2025, after which the cluster underwent several months of testing and was expanded with additional components. It now forms a key part of the CLARA Testbed, the distributed research infrastructure of the CLARA Centre of Excellence, which brings together CIIRC CTU, INDRC, VSB – Technical University of Ostrava and the International Clinical Research Center (ICRC) in Brno with leading European partners, including the Paris Brain Institute (PBI) and the Leibniz Supercomputing Centre (LRZ) in Munich.
“The cluster allows us to approach the challenge of neurodegeneration from two sides. On the one hand, we are developing AI models that allow us to study, at large scale, the behaviour of proteins associated with Alzheimer’s disease and to search for molecules that could influence them. On the other, we are developing methods to teach robots to understand human movement and assist people who need support. This computing infrastructure allows us to achieve results that would otherwise be beyond our reach,” says Prof. Josef Šivic, who leads research at CIIRC CTU within the CLARA Centre.
Understanding disease: modelling proteins and searching for effective compounds
One of the research directions focuses on the development of computational methods for molecular biology. CIIRC CTU is developing these methods in cooperation with ICRC in Brno, which has long been studying apolipoprotein E, a protein whose ApoE4 variant is the strongest known genetic risk factor for Alzheimer’s disease. Using the new computing cluster, the teams are jointly searching for compounds that could prevent the pathological aggregation of the ApoE4 protein.
The research combines two approaches that have traditionally been used separately: accurate but computationally demanding physics-based calculations and fast AI models. The result is a tool that can rapidly and reliably screen large libraries of molecules representing potential drug candidates. Selected molecules are subsequently tested experimentally in ICRC laboratories in Brno. In parallel, generative models are being developed to design entirely new molecules directly within the binding site of a target protein. This approach is intended, in a subsequent phase, to focus specifically on apolipoprotein E.
Artificial intelligence for robotics
The second major area enabled by the new infrastructure is the development of AI methods for robotics. Training large-scale robotic models is among the most computationally demanding tasks in the field and is practically inaccessible without appropriate computing infrastructure.
CIIRC CTU teams are using the cluster, among other things, to work on so-called world models – models that allow a robot to anticipate what will happen in a scene after it takes an action and use this prediction to make decisions. They are also working on methods for teaching large robotic models new skills without causing them to lose abilities they have already learned.
In the future, these methods are expected to support the development of assistive technologies. As part of the CLARA project, CIIRC CTU is modernising its assistive technology laboratory, which includes a system for comprehensive human movement analysis as well as an assistive apartment – a model home where researchers can test how technologies can support people affected by neurodegenerative disorders in everyday environments.
The infrastructure will also support other research directions pursued by CIIRC CTU within the CLARA Centre, including advanced analysis of brain imaging data, led at the institute by Milan Němý, and the analysis of brain electrophysiological signals developed by a research team led by Václav Křemen from the US-based Mayo Clinic.
Technical solution
In recent years, research teams have faced rapidly growing demands for processing biological, imaging and movement data. Existing computing infrastructure was no longer able to meet these demands in terms of performance, scalability and operational reliability. The cluster was delivered as a fully integrated solution by M Computers in cooperation with Hewlett Packard Enterprise following a public tender. Its architecture was specifically designed to meet the needs of the CLARA Centre, combining high computing performance with efficient management and the possibility of future expansion.
“Our goal was not simply to deliver high performance, but to create a stable and sustainable platform capable of growing alongside the needs of scientific research. The combination of NVIDIA GPU technologies, HPE enterprise infrastructure and an open-source software ecosystem represents an exceptional solution for the Czech academic environment,” says Vojtěch Šubrt, Sales Manager at M Computers.
CLARA GPU cluster at CIIRC CTU in numbers
- Computing performance: up to 260 PFLOPS (AI FP8/FP16)
- GPU memory: 9 TB
- GPU accelerators: 48 × NVIDIA H200 and 48 × NVIDIA L40S
- Storage: HPE ClusterStor E1000, with the Lustre parallel file system combining HDD and all-flash storage with a total capacity of 3.5 PB and read/write throughput of up to 250/200 GB/s
- Network: InfiniBand NDR 400 Gb/s
- Power consumption: up to 110 kW
The equipment was acquired with funding from the Jan Amos Komenský Operational Programme (OP JAK) as part of the CLARA project – Centre for Artificial Intelligence and Quantum Computing in System Brain Research (project registration no. CZ.02.01.01/00/23_029/0008437).