Deploying generative AI in security-sensitive operational environments using AI OS ‘Runway’
Building on a partnership that began in 2024 to create a blueprint for AI deployment across public infrastructure
SEOUL, South Korea, June 18, 2026 — MakinaRocks, South Korea’s leading Physical AI company, today announced a joint project with Korea Water Resources Corporation (K-water) to integrate generative AI into K-water’s smart water treatment operations. The project aims to establish a practical and secure AI deployment model for critical public infrastructure.
The project centers on integrating generative AI into K-water’s existing AI-powered autonomous operations system for smart water treatment plants, powered by MakinaRocks’ AI Operating System (AI OS), Runway. The integration is designed to help field operators better understand AI-generated recommendations and make faster, more informed decisions by accessing real-time operational data and internal documents through a single interface.
K-water currently operates AI-powered autonomous systems across 44 water treatment plants nationwide, covering the full water treatment process—from water intake and chemical dosing to coagulation, sedimentation, filtration, disinfection, ozonation, and GAC filtration. Despite the scale of deployment, field operators have faced challenges in understanding the reasoning behind AI recommendations, limiting trust and operational adoption. This project is intended to address that gap.
MakinaRocks will provide an integrated generative AI environment combining large language models (LLMs), retrieval-augmented generation (RAG), and explainable AI (XAI). The system will deliver natural language explanations of AI decisions, enable operators to reference internal documents such as operational manuals and technical diagnostic reports in real time, and visualize how individual process variables contributed to each AI prediction. These capabilities are designed to improve interpretability and support more effective use of AI in day-to-day operations.
A key aspect of the project is its architecture for security-sensitive environments. The system is designed to operate independently of external internet and cloud infrastructure, allowing generative AI to run within K-water’s secure operational environment. MakinaRocks will deploy Runway on GPU servers at K-water’s regional basin offices, providing real-time access to operational data and internal documents without reliance on external connectivity. The project is intended to serve as a replicable model for AI deployment in public infrastructure settings where both security and operational continuity are required.
The partnership builds on collaboration that began in 2024, when MakinaRocks joined K-water’s open digital water platform, WaterRound, and participated in the “MLOps-based Data Analytics Model Operations Framework” initiative. The two organizations have since expanded their work to include AI education programs for WaterRound partner companies and AX-focused data analytics projects, addressing the practical challenges of AI deployment in public-sector environments.
“The water industry presents some of the most demanding conditions for operational AI, where real-time data requirements, strict security constraints, and high-stakes decision-making converge,” said Andre S. Yoon, CEO of MakinaRocks. “Together with K-water, we aim to demonstrate that generative AI can operate securely and reliably in critical infrastructure and deliver measurable value in day-to-day operations.”
