DEEPX, Doosan Robotics Team Up on AI Welding Solution for Nuclear Power Equipment Manufacturing
DEEPX and Doosan Robotics, with SAIGE and Changwon National University, are developing an AI welding solution that pairs cobots with on-device NPUs.
- Selected for the Ministry of Trade, Industry and Resources' Robot Industry Technology Development Program, with SAIGE and Changwon National University joining a four-party consortium
- Intelligent welding solution combines AI, collaborative robots and on-device NPUs
- Partners target quality validation for nuclear power equipment in 2029 and commercialization from 2030
Oct. 6, 2026 — DEEPX, a developer of ultra-low-power AI semiconductors, is teaming up with Doosan Robotics to develop an AI-powered intelligent welding robot solution aimed at easing a shortage of skilled welders in nuclear power plant equipment manufacturing.
The collaboration follows the project's selection for the Ministry of Trade, Industry and Resources' 2026 Robot Industry Technology Development Program. Doosan Robotics, DEEPX, SAIGE and the Changwon National University Industry-Academic Cooperation Foundation signed a memorandum of understanding (MOU) on July 21 to carry out the project and set up a framework for joint technology development.
At the heart of the project is combining vision AI with collaborative robots, or cobots, to make welding smarter. The partners will develop technology that uses AI to recognize and correct welding targets and paths, and that analyzes vision and thermal imaging data collected during welding to assess weld condition and detect defects. They plan to pair this with cobots to create an intelligent welding solution ready for real factory floors.
Doosan Robotics, the lead research institution, will oversee the consortium and design the system integration architecture. It will develop a cobot welding cell, AI that corrects for fine displacement and real-time welder control technology. DEEPX will develop an on-device edge AI unit and software development kit (SDK) built on its Korean-made neural processing unit (NPU). It will also build a three-channel vision data pipeline, integrate with the robot controller, develop AI models for thermal image analysis, and support environmental durability testing and reliability certification for welding environments.
SAIGE will develop AI models for 3D vision-based recognition of welding targets, weld path correction and weld bead defect detection. The Changwon National University Industry-Academic Cooperation Foundation will develop AI-based weld path generation and physics-based digital twin technology.
The four organizations will integrate their technologies into a single intelligent welding system and carry out phased field testing at Doosan Enerbility's nuclear power equipment manufacturing site during the project. They aim to complete quality validation for nuclear power plant equipment in 2029 and begin commercialization in 2030.
Supported by the Ministry of Trade, Industry and Resources and the Korea Evaluation Institute of Industrial Technology (KEIT), the project links research and development with field testing to bring AI and cobot technology into actual manufacturing processes.
"This project will be a concrete example of physical AI, where ultra-low-power AI chips are combined with robots and used on real manufacturing floors. Building on DEEPX's NPU and on-device AI technology, we will help create an AI computing environment that handles the AI workloads manufacturing sites demand efficiently and reliably." — DEEPX
"Through this collaboration, we plan to combine Doosan Robotics' cobot technology with our partners' AI technology to develop an intelligent welding solution that manufacturers can actually use. We will help address the challenges manufacturing sites face, such as a shortage of skilled workers, and expand automation and intelligence in welding." — A Doosan Robotics official



