Top Data Science received an exciting opportunity to use the LUMI supercomputer
Top Data Science was selected as a partner in the AISA project led by Nokia and financed by Business Finland. In the project, they were offered the opportunity to use one of the world´s fastest supercomputers, EuroHPC LUMI. Supercomputers had previously only been available to universities and research organizations.
Top Data Science specializes in artificial intelligence, machine learning, and software engineering services. The majority of its customers are in Nordic countries, Germany, and Japan. The biggest solution segment is based on computer vision technologies, which utilize video and other image data to collect information relevant to the customer use case in focus.
“With our artificial intelligence-based services, our customers can, for example, establish more automated factory production and quality assurance processes. Many times these include multiple simultaneous tasks such as visual inspections, checking measurements and product markings,” says Kai Lehtinen, COO at Top Data Science.
“Another part of our business is developing and implementing business and process optimization solutions. We have the ability to predict process behavior based on historical and real-time sensor data. In process optimization, the customer objective is, for example, to achieve desired product quality with minimal use of raw materials, chemicals, and energy. In business optimization, the customer objective is to maximize operating profit. We enable this by helping them to create optimal production and sales plans,” he adds.
Combined AI and supercomputing knowledge a plus
Top Data Science was chosen as a partner in the AISA (AI-based Situational Awareness) project, funded by Business Finland and led by Nokia. This project aims to enhance applying AI and high-speed wireless communications among Finnish industrial companies and research institutes. Top Data Science stands out as the sole company specializing in AI within the consortium. Business Finland offered Top Data Science an opportunity to utilize the LUMI supercomputer.
Top Data Science was naturally excited about the opportunity to experiment with LUMI usage in their R&D. The decision to utilize the opportunity was supported by Lead Data Scientist, Kseniia Khakalo’s prior experience with supercomputers.
“I worked with the Triton HPC cluster at Aalto University. It also uses the Slurm Workload Manager so I was familiar with the concept of submitting jobs, storing results, and downloading data,” says Kseniia. This prior knowledge eased their adoption of LUMI in industrial projects.
Scalability, ease of use, and good documentation
Top Data Science leveraged LUMI for human pose estimation analysis, processing hundreds of video clips in parallel to create a dataset of human skeleton trajectories.
“Overall, LUMI provided a powerful and effective platform for our projects, meeting requirements, and facilitating workflow,” explains Kseniia. “We were able to effectively outsource computationally heavy tasks and generate the desired human skeleton trajectory dataset, which is now actively utilized in our work.”
LUMI’s capability to run parallel jobs on GPUs and its flexible containerization approach offered the needed scalability. Additionally, its comprehensive documentation with plenty of examples and easy navigation allowed them to find the necessary information without extra support.
“I was pleasantly surprised by how smoothly everything went. The documentation was a great resource; LUMI exceeded my expectations in terms of functionality and ease of use,” adds Kseniia.
Advantages of using LUMI
“The primary advantage for us is to understand what LUMI offers, and it should be strategically positioned for our AI solution development. This helps us address customer problems and deliver tailored solutions that go beyond what the normal computational setup offers. It’s crucial to understand LUMI’s strengths and limitations, such as its suitability for inference tasks. Comprehensive documentation and containerization features make it an asset for various projects,” says Kai Lehtinen.
LUMI’s advanced capabilities can enhance workflow and drive innovation, provided its resources are leveraged effectively.
It is important for CSC to understand the R&D needs of large industrial AI initiatives such as AISA, and, in this respect, the Top Data Science project has given us valuable experience for the future,” says Dan Still, Partnerships Manager at CSC.
More information about LUMI’s business use is available on our website in section High-performance computing for companies.
Dan Still
Dan Still works with building industrial partnerships and networks to boost industrial HPC use.
