• Lablup's sixth annual technology conference to be held September 29 at COEX, Seoul
• Themed "Make AI Composable²," the event spotlights the end-to-end AI ecosystem, from infrastructure to inference, agents, and security
• Lablup to unveil Continuum Router, a control plane for LLM monitoring and billing, and Continuum Hub, a unified management layer
• Leading AI companies and institutions including NVIDIA, kt cloud, FuriosaAI, Samsung Welstory, Upstage, and Nota AI to participate

SEOUL, South Korea — Lablup (CEO Jeongkyu Shin), an AI infrastructure management and operations software company, today announced that it will host its annual technology conference, "lab | up > /conf/6," on September 29 in the Conference Room on the 3rd floor of COEX in Samseong-dong, Seoul.

Now in its sixth year, the Lablup Conference is where Lablup shares its AI infrastructure technology and R&D achievements, and where industry professionals and developers come together to discuss technical challenges and their solutions. Since launching online in 2021, the conference has expanded its scope each year to keep pace with major technological shifts in the AI industry.

This year's theme is "Make AI Composable²." Building on last year's "Make AI Composable" concept, it presents a vision for AI infrastructure that lets organizations select and combine rapidly diversifying AI models, semiconductors, infrastructure, data, and agent technologies as needed, while operating them seamlessly within a single environment.

The AI industry's technical demands are expanding quickly, moving beyond building GPU infrastructure for large-scale model training to inference performance and cost efficiency in production, the use of diverse AI accelerators, multi-cloud operations, and AI agent management and security. As AI service usage grows, the ability to connect a wide range of models efficiently and to manage token usage systematically has become an increasingly critical part of infrastructure operations.

Jeongkyu Shin, CEO of Lablup, will deliver the opening keynote, "On End-to-End in the Era of the AI Industry." As AI rapidly spreads from development into industries and services, he will discuss the need for end-to-end AI platforms that connect the entire process, from building infrastructure to model training, inference, and service operations, and share Lablup's direction.

The keynote will be followed by a panel talk titled "Inference Optimization to Break the AI Bottleneck." Joongi Kim, CTO of Lablup, will be joined by Jeongseok Noh, CEO of BFactory; Tae-Ho Kim, CTO of Nota AI; Jinwon Lee, CTO of HyperAccel; and Hwalsuk Lee, CTO of Upstage, to discuss how to improve inference performance and infrastructure efficiency, which are growing in importance as AI services scale.

Lablup will also introduce Continuum Router and Continuum Hub, which connect multiple AI models through a single gateway. In his session, Jongmin Kim, Go-to-Market Lead at Lablup, will explain Lablup's concept of the "Token Factory," focusing on technology that flexibly routes requests across diverse AI models and centrally manages token usage. Attendees will also be able to see Continuum Router and Continuum Hub in action through live demos at the venue.

The conference will cover a broad range of topics, including AI infrastructure, inference and serving, AI agents, platform engineering, security, data pipelines, and physical AI. Sessions will go beyond technology trends to address the challenges that arise in building and operating real-world AI systems, their solutions, and case studies from the field.

External speakers from NVIDIA, VAST Data, kt cloud, FuriosaAI, Samsung Welstory, RLWRLD, AIM Intelligence, and KAIST will take the stage. Topics will include system acceleration in the age of AI agents, core principles for building an AI factory, composable platform design, high-performance LLM serving on FuriosaAI's RNGD, AI agent workflows in enterprise environments, AI agent security, and efficient LLM serving for multi-agent systems — all addressing the technical challenges of adopting and operating AI in practice.

"As the AI industry grows rapidly, how well you connect diverse models and infrastructure, and how efficiently you manage the tokens they generate, is becoming a key competitive advantage," said Jeongkyu Shin, CEO of Lablup. "At lab | up > /conf/6, we will show what 'Make AI Composable²' means and how the 'Token Factory' comes to life in real-world AI operations."

"lab | up > /conf/6" will take place on September 29 from 9:30 a.m. to 5:40 p.m. KST in the Conference Room on the 3rd floor of COEX, Seoul.