October 15, 2021
With more and more enterprises turning to AI for a myriad of tasks, companies quickly find out that training AI models is expensive, difficult and time-consuming. Finding a new approach to deal with those cascading challenges is the aim of a new startup, MosaicML, that just came out of stealth... Read more…
Five Recommendations to Optimize Data Pipelines
When building AI systems at scale, managing the flow of data can make or break a business. The various stages of the AI data pipeline pose unique challenges that can disrupt or misdirect the flow of data, ultimately impacting the effectiveness of AI storage and systems.
With so many applications and diverse requirements for data types, management systems, workloads, and compliance regulations, these challenges are only amplified. Without a clear, continuous flow of data throughout the AI data lifecycle, AI models can perform poorly or even dangerously.
To ensure your AI systems are optimized, follow these five essential steps to eliminate bottlenecks and maximize efficiency.
Karlsruhe Institute of Technology (KIT) is an elite public research university located in Karlsruhe, Germany and is engaged in a broad range of disciplines in natural sciences, engineering, economics, humanities, and social sciences. For institutions like KIT, HPC has become indispensable to cutting-edge research in these areas.
KIT’s HoreKa supercomputer supports hundreds of research initiatives including a project aimed at predicting when the Earth’s ozone layer will be fully healed. With HoreKa, projects like these can process larger amounts of data enabling researchers to deepen their understanding of highly complex natural processes.
Read this case study to learn how KIT implemented their supercomputer powered by Lenovo ThinkSystem servers, featuring Lenovo Neptune™ liquid cooling technology, to attain higher performance while reducing power consumption.
© 2023 HPCwire. All Rights Reserved. A Tabor Communications Publication
Reproduction in whole or in part in any form or medium without express written permission of Tabor Communications, Inc. is prohibited.