KACI Holds Technology Development Results Sharing Event for AI-PaaS and AI-MDC

The Korea Artificial Intelligence Cloud Industry Association (KACI) held the "PaaS Technology Development Results Sharing Event for AI Model and Service Development" and the "AI-MDC Technology Development Project Results Sharing Event."

According to an announcement on the 2nd, the Korea Artificial Intelligence Cloud Industry Association (KACI) held the two results-sharing events to inform the industry about technology development achievements related to AI-PaaS and AI-MDC.

The event was organized to present research results on a cloud-native-based AI development platform and distributed infrastructure based on domestic AI semiconductors, as well as to share a range of field demonstration results.

It was also arranged to expand the use of the technologies and broaden the basis for business collaboration.

Among them, the "AI-PaaS Technology Development Results Sharing Event" was promoted with support from the Ministry of Science and ICT and the Institute for Information & Communications Technology Planning & Evaluation (IITP), and was aimed at disseminating the results of the R&D project. The project sought to secure core AI-PaaS technologies capable of responding to various cloud service provider (CSP) environments and enabling the rapid development and deployment of AI models and services based on cloud-native technologies, as well as flexible linkage with external systems.

The "AI-MDC Technology Development Project Results Sharing Event" took place with the participation of the lead organization, Innogrid, and partner organizations Theramind, the Korea Institute of Science and Technology Information (KISTI), BioBrain, SoftOnNet, Polaris Office, and the Korea Artificial Intelligence Cloud Industry Association. The event was arranged as a venue to introduce platform technologies for end-to-end support of AI services.

At the event, an AI-PaaS integrated gateway and portal supporting model registration and execution, an integrated monitoring system for AI model development workflows, and a data lakehouse were presented. The technologies introduced were based on the scope of platform technologies for end-to-end support of AI services.

Actual demonstration results were also disclosed. The unveiled results included BioBrain's demonstrations in the vaccine and new drug fields, SoftOnNet's vision AI service for logistics transport asset management, and Polaris Office's document writing and editing support technology.

These results were presented as examples applied to the biotech, logistics, and document industries. Through the disclosure of such real-world demonstration outcomes, the event highlighted the technologies' potential for industrial application.

This "AI-MDC Technology Development Project Results Sharing Event" was held as a venue to share rack-level all-in-one micro data center technology equipped with domestic AI semiconductors. The photo shows the site of the "AI-MDC Technology Development Project Results Sharing Event," courtesy of KACI.

The project aims to move away from reliance on a single large centralized data center and focuses on building a distributed infrastructure by placing edge computing resources near demand points for AI services, such as manufacturing sites, hospitals, and public facilities, and then integrating and operating them through the cloud. The goal is to foster a practical service ecosystem for domestic AI semiconductors.

The project is overseen by the Electronics and Telecommunications Research Institute (ETRI) and is being promoted by linking four detailed subprojects. The subprojects consist of micro data center infrastructure building, scale-out cloud integration centered on Innogrid, operating software development, and AI service demonstrations.

At the event, the "AI Service Open Platform" technology was shared. The technologies shared include the heterogeneous domestic AI semiconductor-supporting development tool (u-SDK), multi-model optimization pipelining, a user portal based on a large language model (LLM), and cloud integration technology for distributed resource unified management and scaling.

In addition, results from demonstrations in three major industrial settings were introduced. The demonstrations included a home training service for respiratory rehabilitation in the medical field, a shipyard collaborative autonomous mobile robot (AMR) in manufacturing, and intelligent CCTV control based on a vision language model (VLM) and a small large language model (sLLM) in the public sector.

On the strength of these results, the research institutions and association involved plan to accelerate technology advancement and industrial expansion. In the next phase, the participating organizations and the association plan to focus on further advancing the technologies and broadening their range of industrial applications.

The AI-PaaS research team will continue technical cooperation among participating organizations and pursue expansion into areas where the technologies can be applied to real services. It will also fully commercialize the research results through Innogrid's AI and MLOps platform, "AICubeit." The AI-MDC research team plans to improve technical completeness by reflecting demonstration data from manufacturing, medical, and public-sector sites. In addition, it plans to open operating software and development environments to external users and gradually expand the scale of domestic NPU integration.

The Korea Artificial Intelligence Cloud Industry Association plans to actively identify companies with technology demand by leveraging its own network. The association will continue technology exchanges and promotion of results, and will keep supporting efforts to ensure that the developed technologies lead to actual business outcomes in the industry. It also plans to provide comprehensive support.

Source: IT DAILY · Kwon Young-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=241992

References

• https://www.itdaily.kr/news/articleView.html?idxno=241992