ISC 26 | KAYTUS

Event Information

ISC High Performance is an annual global event that connects users, developers, and providers of HPC technology, which is celebrating its 40th anniversary this year from June 22-26 in Hamburg, Germany.

KAYTUS is proudly attending ISC26 as a Gold sponsor, and we are delighted to invite you to visit us at booth #D19, where we will be presenting our latest technological advancements tailored for the AI and HPC sectors: our latest HPC innovations, cutting-edge server products, and advanced liquid-cooling and AIDC solutions. To make your experience even more engaging, we have prepared interactive challenges with exciting giveaways designed to highlight our latest technologies.

Highlight

High-Density AIDC Solutions

KAYTUS is committed to achieving cost parity between air and liquid cooling by promoting the scaling and industrialization of liquid cooling. Discover how we improve production efficiency, drive economies of scale, lower adoption barriers for liquid cooling among data centers, and support green transformation across industries.
End-to-end AI&HPC Solutions

KAYTUS is at the forefront of innovation with its DevOps Al platform for enterprise, MotusAl, and the Brilliant Computing System (BCS). Learn how these platforms deliver full control and transparency over AI infrastructure, workloads, and user management.
Future-Ready V3 Servers

V3 server lineup delivers extreme performance, energy efficiency, and scalability. Engineered to seamlessly support cloud, big data, and performance-driven workloads and applications, it accelerates innovation while reducing operational costs. Embrace the future of sustainable, high-performance computing.

Expert Talk

Agent 4 Science: Bringing AI Agents into Scientific Workflows and HPC Simulation

Time:10:40-11:00am, June 25

Location: HPC Solutions Forum, Hall H Booth L01, Congress Center Hamburg

Speaker: Dr. Di Wang

Large language models have demonstrated remarkable capabilities in code understanding and generation, yet theirintegration into HPC and scientific computing workflows remains largely unexplored. The gap is not in what LLMs knowabout optimization -they possess broad knowledge of compiler flags, vectorization, memory hierarchy, and algorithmictechniques but in how to let them act reliably on real codebases with real performance constraintsIn this talk, we present our experience building and iterating on an LLM-driven agent system for automated HPOperformance optimization. The agent autonomously profiles applications, forms structured hypotheses aboutperformance bottlenecks, applies source-level and build-level optimizations, and verifies results through statisticalyrigorous benchmarking all while maintaining a separation between immutable experimental facts and the agent'sevolving interpretations.