• Joongi Kim, CTO of Lablup, presents "Building the AI Factory with Inference and Routing Optimization and Token Monitoring"
• Lablup shares AI infrastructure operations strategies, including inference and routing optimization and token monitoring, as AI services scale
• Building on joint GPUaaS development with KT, the two companies continue to expand technical collaboration and the AI infrastructure ecosystem

SEOUL, South Korea, — Lablup (CEO Jeongkyu Shin), an AI infrastructure software company, participated in "kt cloud summit 2026," held on September 15 at the Grand InterContinental Seoul Parnas, where it presented strategies for inference and routing optimization and token monitoring for AI factory operations.

The kt cloud summit is kt cloud's flagship annual event, where the company shares the technologies, industry trends, and real-world use cases needed for AI transformation (AX), with a focus on AI data centers (AIDC) and the cloud.

Under the theme "AX for Your Business: Easier AX Execution, Faster Business Growth," this year's summit showcased a wide range of technologies and case studies spanning the full scope of building and operating AI services, including AI data centers, cloud platforms, AI factories, disaster recovery (DR), and security.

Representing Lablup, CTO Joongi Kim spoke in the "AIDC & AX Infrastructure" track on "Building the AI Factory with Inference and Routing Optimization and Token Monitoring."

In his presentation, Kim introduced approaches to routing optimization for efficiently handling the inference requests generated in AI service operations, along with methods for monitoring token usage. He outlined how these approaches improve the operational efficiency of AI infrastructure and shared Lablup's direction for implementing AI factories that can run and scale large-scale AI services reliably.

Lablup also continues to collaborate with KT in AI infrastructure. Last year, the two companies signed a memorandum of understanding (MOU) to jointly develop a GPU cloud service (GPUaaS), working together to build services that help enterprises use high-performance GPU resources more efficiently. Going forward, the two companies plan to continue expanding their collaboration in AI infrastructure.

"As the center of gravity in AI infrastructure shifts rapidly from large-scale training to inference and real-world service operations, efficiently handling requests across diverse models and users has emerged as a new challenge," said Joongi Kim, CTO of Lablup. "At kt cloud summit, we shared the experience Lablup has built in AI infrastructure operations and inference optimization. We will continue to help enterprises build and scale their AI services more efficiently."