Xperix Builds AI Full-Stack Based on Furiosa AI NPU for Public AX Project View original image

Xperix is accelerating its AI full-stack business that applies domestically produced artificial intelligence (AI) semiconductors to real work in public institutions. The company is targeting the public sector by incorporating FuriosaAI's homegrown neural processing unit (NPU) into the government-driven Public AI Transformation (AX) project, spearheaded by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA).


On September 15, Xperix, an AI solutions specialist, announced that the company has established a sovereign AI full-stack system based on FuriosaAI's second-generation AI accelerator, RNGD (Renegade), by participating in the 'Public AX Project' under the education and training division.


This project is a case of implementing an AI full-stack—combining a Korean-made NPU and an independently developed foundation model—in the actual operations of public organizations. System construction, proof-of-concept, and on-site deployment are currently underway.


The Public AX Project is designed to discover and expand public sector AI transformation cases that citizens can tangibly experience, delivering them to various government departments through public-private partnerships. In the education and training division, AI is leveraged to detect suspected cases of fraudulent course attendance and improper benefit claims within the curriculum managed by the Korea Industrial Human Resources Corporation, and to support related administrative work.


Xperix is involved in every stage of the project—from constructing and analyzing training data, to developing and validating AI solutions, building on-premises environments, and deploying solutions in the field.


Tangible improvements in real work efficiency are already being seen. A representative from the Korea Industrial Human Resources Corporation stated, "Xperix is engaged throughout the process of building AI solutions for fraudulent training detection and operational support. By applying AI to actual work, data analysis and validation on fraudulent claims—which previously took two to three months—have been reduced to less than two weeks, along with collaborative work between relevant organizations."


The representative continued, "Through chat-based AI, complex queries about attendance, enrollment, and connection records are now easier to handle, while evidence of suspected improper benefits, abnormal patterns, as well as items requiring further verification and requested materials, can all be identified quickly. AI is evolving beyond a simple data analysis tool and is now being used as a practical work support system, assisting field personnel in decision-making and facilitating collaboration across government agencies."


Looking ahead, it is expected that, by leveraging accumulated data, the AI can continuously discover new patterns of improper claims and proactively detect them—connecting seamlessly to field investigations—thus shifting the focus from post-incident detection to prevention-oriented work.


The core inference infrastructure of this system utilizes FuriosaAI's RNGD. For some tasks, Xperix's own foundation model has been integrated. One of the unique aspects is connecting AI infrastructure, models, data, platform, and application services into a single system in an on-premises environment, ensuring that sensitive public data is processed internally without being transmitted externally.


Xperix is establishing an end-to-end AI full-stack business system that provides six stages—AI consulting, infrastructure, platforms, software, maintenance, and customer success—through a single window. This single-source approach from design to operation shortens both the build period and response time to disruptions.


Bokun Park, Head of Business Division at Xperix, commented, "This is a leading case of building and deploying a full-stack AI system for real public services by combining a domestic NPU and sovereign AI model. By applying it to actual public work, we aim to demonstrate our capabilities to build a comprehensive end-to-end AI full-stack—from AI semiconductors to infrastructure and services."



In addition to maintaining its partnership with FuriosaAI, Xperix recently merged with AI solutions provider Ainut. Building on this, the company plans to expand its participation in national, public, and regional AX projects by supplying NPU appliance-based AI inference and on-device AI, as well as providing systems integration (SI)-based AI hybrid infrastructure solutions.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing