Hewlett Packard Enterprise
Nanya accelerates the use of AI in production
Nanya, the world's fourth largest manufacturer of DRAM, uses artificial intelligence to increase productivity and quality in production. Now, with the help of Hewlett Packard Enterprise, the roll-out of new AI applications in the semiconductor plants has been significantly accelerated.
For Nanya Technology Corporation (Nanya), AI and advanced analytics are strategic tools for detecting errors, preventing failures and increasing the level of automation in its production. In recent months, Nanya has significantly increased the number of AI use cases. This has also increased the complexity of the AI environment. Nanya introduced the 'Ezmeral Container Platform' from Hewlett Packard Enterprise (HPE) to centralize the management of this environment and accelerate the rollout of new AI use cases.
AI use cases at Nanya include, for example, video quality control of DRAM components and predictive maintenance of production machines. Multiple data teams are driving dozens of AI projects in parallel.
To set up a new AI model, Nanya's data scientists must implement a combination of tools, frameworks, data sources and graphics processing unit (GPU) systems tailored to the specific use case. Previously, this could take days or weeks because the tools had to be installed manually, and because the data had to be copied and transferred to the appropriate GPU systems.
With the 'Ezmeral Container Platform', the Nanya teams now have uniform access to all tools, systems and data sources. The multi-client capability of the platform ensures a logical separation between the projects. Data no longer needs to be copied and transferred, as the HPE platform enables secure remote access. The platform also includes an app store with tools and frameworks for AI and analytics. Nanya's data scientists can select these tools from the app store and install them with a click of the mouse. As a result, they can set up new AI use cases in just a few minutes. This allows the training of AI models to begin earlier, so that AI applications can be used in production more quickly.










