On October 1, 2026, we welcomed nearly 150 participants to the modern premises of the Diplomat Hotel in Prague for the annual MetaCentrum High-Performance Computing Seminar 2026. This year, it took place as a separate full-day track within the larger two-day CESNET 2026 Conference. The event confirmed how rapidly research computing is evolving. Today, traditional HPC is accompanied by the processing of large datasets using cloud technologies and the integration of artificial intelligence, including large language models (LLMs).
News, Vision, and Research Data Management
The introductory word was given by the head of the MetaCentrum activity Miroslav Ruda, who introduced the participants to the current changes in the MetaCentrum service portfolio and presented the strategic outlook for the upcoming period. He opened the key topic of effective management and long-term storage of scientific data.
David Antoš, head of the CESNET data storage department, followed up with a presentation of data services, national storages, and repositories, which are an absolute necessity today for safe data storage and sharing. Practical experience on this topic was subsequently added by Tomáš Svoboda (CEITEC). He presented the involvement of the OneData system and shared real experience with deploying the S3 interface directly in e-INFRA CZ.
The morning infrastructure block was later complemented by Klára Moravcová with an introduction to the latest news in the flexible MetaCentrum Cloud environment.
Infrastructure, Orchestration, and Sustainability
This year, a lot of space was given to advanced orchestration and management of computing capacities. Lukáš Hejtmánek (CERIT-SC) analyzed the issue of Kubernetes and the Rancher platform in detail. He showed how these tools can be used maximally effectively not only for running one's own services but also for the research applications themselves, which operate today "in the shadow of AI".
The very current topic of sustainability and the ecological impacts of infrastructure was presented by Jiří Sitera. In his lecture on Green Computing, he presented the international GreenDIGIT project. It helps data center administrators reduce the environmental footprint of demanding IT operations while preparing them for extreme climate events such as summer heat waves.
Dalibor Klusáček and Václav Chlumský spoke about optimizing the normal operation of the infrastructure from the perspective of the users themselves. They explained to the audience the detailed functioning of the job scheduler and the "fairshare" priority mechanism. They also added useful procedures on how researchers can adjust the settings of their jobs to achieve the best possible efficiency.
User Tools and Practical Experience
The afternoon block showed that modern research computing is increasingly moving from complex command lines to more intuitive web interfaces. Martin Čech demonstrated the capabilities of the Galaxy platform, which allows for convenient composition and analysis of data workflows directly in the web browser. He was followed by Aleš Křenek with an introduction to the MDDash environment – a specialized all-in-one virtual environment designed for molecular dynamics simulations.
Examples directly from laboratories and user experience were brought by Martin Krupička (UCT Prague), who captivated with a lecture on transferring his jobs from the standard PBS scheduler to the Kubernetes container environment using the ClaudeCode AI assistant, and Filip Jozefov (IOCB Prague), who discussed the methodology and obstacles in the detection of small molecules compared to protein research.
The user section was summarized by Jiří Vorel, who officially introduced the new supportive Ambassador Program, which aims to ensure even closer and more targeted support for researchers and the most important operational news.
AI, LLMs, and Development with Autonomous Agents
Just like last year, the conclusion belonged to artificial intelligence. The final block was opened by Lukáš Hejtmánek with a lecture on AI subtitled "Hello Computer", in which he focused on the operation of large AI models (LLMs). Participants learned how they can be integrated into their own development via WebUI or API and how to use robust Kubernetes orchestration to handle even the most demanding AI tasks.
Terézia Slanináková then outlined the current possibilities of application development using autonomous AI agents in her presentation. She gave participants valuable advice from real deployments and showed concrete scenarios of how teams can speed up and facilitate software development in practice thanks to AI agents.
Presentations for Download and Video Recordings
You can view and download the complete program and all the discussed presentations on the official Seminar 2026 page.
If you could not attend the event in person, a recording of the entire day is available on YouTube — you can easily play the individual lectures in this playlist (with EN subtitles available).
