J&Tech and Pliops Partner to Deliver GenAI/LLM Acceleration Solution to the Korean Market

J&Tech and Pliops announced a strategic partnership to offer comprehensive large language model (LLM) acceleration solutions in Korea.This collaboration combines J&Tech’s deep expertise in infrastructure technology with Pliops’ innovative GenAI LLM inference acceleration product, LightningAI.

Under the terms of the agreement, J&Tech will supply and integrate infrastructure systems optimized for GenAI workloads, and Pliops will contribute its LightningAI product. LightningAI optimizes GenAI inference infrastructure by reducing GPU compute per chat query I/O, unlocking seamless data retention, cost savings, and breakthrough performance for GenAI and LLM applications. Built as an Accelerated Key-Value distributed smart node, LightningAI introduces a new tier of memory below HBM for GPU compute workloads. It accelerates end-to-end performance by more than 3X and enhances efficiency across all leading inferencing frameworks, including vLLM and others. Leveraging Pliops’ state-of-the-art technology, LightningAI substantially improves GPU utilization, enabling significant TCO reductions, lower power consumption, and decreased carbon emissions – delivering advanced AI solutions for the next gen of intelligent infrastructure.

“We are proud to partner with Pliops to bring a unique GenAI acceleration solution to Korean enterprises,” said Mr. Rha, CEO, J&Tech. “This collaboration brings together the strengths of both companies to address critical GenAI performance and efficiency challenges, giving Korean enterprises the ability to deploy high-performance, locally supported GenAI systems. Together, we’re enabling faster innovation, stronger data sovereignty, and a more competitive AI ecosystem in Korea.”

This collaboration follows an internal proof-of-concept (POC) exercise conducted at J&Tech’s premises, during which the two companies jointly measured end-to-end system performance for vLLM inference. LightningAI demonstrated up to 9x reduction in prefill (TTFT) latency and 3.3x end-to-end throughput (RPS) increase on the open Qwen2.5-32B-Instruct-GPTQ-Int8 model. These results translate into substantial TCO savings and faster deployment of AI-driven services, positioning the solution for commercial rollout in Korea this quarter. The partners are now preparing customer trials with a leading bank and additional enterprise clients in Korea.

“Our acceleration technology delivers breakthrough performance for LLM inference and retrieval – something no other solution on the market can match,” said Ido Bukspan, CEO, Pliops. “By combining this with J&Tech’s infrastructure expertise and deep understanding of the Korean market, we’re making it easier for enterprises to scale GenAI across real-world applications like chatbots, copilots, autonomous agents, and retrieval-augmented workflows.”