Utilization of NVIDIA DGX Systems for Advanced Machine Learning Applications

CEITEC BUT needed to provide researchers and small and medium-sized enterprises with access to high-performance infrastructure for the development and testing of AI applications. Companies often lacked the resources for their own hardware or the know-how to leverage artificial intelligence, which slowed down innovation and industrial digitalization. There was no option to process large volumes of data from manufacturing machines and robots in real time and to test solutions in practice. Therefore, it was crucial to build an accessible platform that would support modernizing production, increase the competitiveness of companies, and accelerate the adoption of AI technologies in both industry and scientific research.

NVIDIA DGX A100 and NVIDIA DGX H100 computing systems were installed, connected via a high-speed InfiniBand network, enabling real-time processing of large-scale data. The solution offers pre-installed and optimized environments for machine learning, including NVIDIA AI Enterprise tools and NVIDIA Base Command for efficient management and deployment of AI applications. Thanks to this, CEITEC BUT provides companies and researchers with 100% subsidized access to testing, education, and development of AI solutions, particularly in the areas of digitization and robotics. The system supports rapid deployment, high levels of automation, and significantly shorter time to innovation.

• Researchers and academics

• Data analysts and AI specialists

• Engineers and technicians from manufacturing companies

• Managers of small and medium-sized enterprises

6

The solution uses the NVIDIA AI Enterprise software stack, which includes optimized environments for machine learning and data analytics. It also incorporates NVIDIA Base Command for efficient infrastructure management and NVIDIA GPU Cloud (NGC) with Docker containers of popular AI frameworks such as TensorFlow, PyTorch, MXNet, Theano, Caffe, and Caffe2. Development takes place in a Docker environment on the DGX OS / Ubuntu operating system, enabling rapid deployment and scaling of AI applications.

C-level executives

Complete realization

internal team

IT manager, Data scientist

The IT infrastructure was the main part of the delivery.

Thanks to the new NVIDIA DGX systems, both researchers and companies gained access to powerful and easily accessible AI infrastructure, which significantly improved the quality of research and industrial projects. The process of testing and deploying AI applications accelerated from weeks to hours, and decision-making was streamlined through real-time data analysis. Companies can now experiment with AI safely and at no cost, which increases innovation, strengthens collaboration between research and industry, and enhances the overall competitiveness of the region.

“In the framework of our EDIH and TEF services, we provide companies with the opportunity to experiment with AI, educate themselves, and test AI applications on state-of-the-art systems that are part of the newly installed supercomputer,” explains Prof. Ing. Pavel Václavek, Ph.D. “This enables small and medium-sized enterprises with up to 499 employees to use advanced technologies at 100% subsidized cost. Our goal is also to integrate the DGX system with other technologies from our RICAIP Testbed Brno, so that we can process data from manufacturing machines and robots in real time.”

Management involvement from the initial project preparation phase through implementation to handover and user training.

Technological integration of multiple generations of NVIDIA GPU systems (Multi-GPU cluster).

Consult with companies that already have experience or references with planned projects and involve them during the planning phase — this helps to properly allocate resources and ensure technical readiness (e.g., data centers) for running the AI infrastructure.