Mobilint NPU support lands in Ray open-source AI framework
Mobilint says Ray can now detect its NPUs automatically and assign resources per AI task, work developed with ETRI.
Mobilint announced on November 4, 2026, that a feature it developed to automatically recognize and manage its neural processing unit (NPU) has been included in the official code of Ray, the open-source distributed AI framework. The company said the change removes the need for manual configuration when its accelerators are used with Ray.
According to digitaltoday.co.kr, the feature was merged into the main branch of Ray's official GitHub repository on August 13, 2026, after official code review. Ray can now detect Mobilint NPUs installed in a system and allocate the resources each AI task needs, and the NPUs can be distinguished and managed per task on servers carrying multiple NPUs or across multi-server environments.
Previously, developers who wanted to use Mobilint NPUs in Ray had to register accelerator resources separately and specify the NPU for each task.
The work is an output of ETRI's project on distributed inference and model optimization technology for heterogeneous AI semiconductors. ETRI principal researcher Kim Young-ju led the project with researchers Kang Min-su, Kang Hyun-seung and Seok Jong-su.
Mobilint said it plans to continue technical cooperation with ETRI and to advance its own software technologies, including its SDK and runtime. The company's hardware line-up includes the ARIES AI accelerator and the REGULUS on-device AI SoC.
Quick answers
What did Mobilint announce?
On November 4, 2026, Mobilint announced that automatic recognition and management of its NPU had been included in the official code of the open-source distributed AI framework Ray.
When was the feature merged into Ray?
It was merged into the main branch of Ray's official GitHub repository on August 13, 2026, after official code review.
Who developed the feature?
It is an output of ETRI's project on distributed inference and model optimization for heterogeneous AI semiconductors, led by principal researcher Kim Young-ju with researchers Kang Min-su, Kang Hyun-seung and Seok Jong-su.