Thaki Cloud Supplies NPU-Based AI Platform to ETRI in Collaboration with OpsNow

SEOUL, South Korea, October 19, 2026 — Thaki Cloud, an enterprise AI infrastructure and platform company, announced that it has supplied its Aegis private cloud platform and Metis AI model serving and inference platform for an NPU (neural processing unit)-based AI research environment at the Electronics and Telecommunications Research Institute (ETRI), in collaboration with cloud operations specialist OpsNow.

The project focuses on combining NPU-based computing infrastructure with a cloud operating environment and AI model execution platform to support the use of accelerator resources and inference workloads for research purposes. Thaki Cloud implemented its software platforms and worked with OpsNow to coordinate the technical and operational elements required for the project.

The research environment uses servers equipped with AMD EPYC processors and Rebellions ATOM-Max NPUs. Thaki Cloud applied Aegis to provide an environment for configuring and managing the underlying infrastructure, and Metis to support NPU-based model serving and inference workloads. The implementation also took into account operational requirements such as API-based integration and resource status monitoring.

Putting AI accelerators, including NPUs, to work in research settings requires more than hardware alone. An infrastructure layer is needed to manage servers, networks, and accelerator resources, alongside an AI platform layer that supports model deployment and inference execution on those resources. Integration and management approaches tailored to the research objectives and system configuration are particularly important.

The project represents an application of Thaki Cloud's private cloud infrastructure technology and AI model serving and inference platform within a single NPU-based research environment. Building on its collaboration with OpsNow, Thaki Cloud plans to address AI infrastructure deployment and operating needs among public research organizations and enterprises.

“Using AI accelerators such as NPUs in research environments requires not only securing the hardware, but also establishing the systems needed to operate resources and run AI models,” said Soryun Kang, CEO of Thaki Cloud. “With Aegis and Metis as our foundation, we will continue developing platform technologies to support a range of AI infrastructure environments.”