Google Cloud Supports POSCO Group to Accelerate AI Innovation... Building a Search Portal

Google Cloud announced on the 5th that POSCO Holdings has developed a generative artificial intelligence (AI)-based material technology and industry trend reporting system and a knowledge search Q&A portal using Google's Gemini 1.5 model. Additionally, it was reported that a pipeline from data to machine learning (ML) was built on its AI platform based on Google Cloud.

POSCO Group has been promoting comprehensive research in core business areas by expanding the development and adoption of cutting-edge technologies such as AI and robotics, centered on the Future Technology Research Institute. In 2022, the AI Research Institute was established to drive innovation in manufacturing processes through strategy formulation and model design, utilizing AI technology for optimizing production processes in traditional manufacturing industries including steel, as well as developing new materials for secondary batteries.

Among them, the Applied AI Research Team developed a generative AI-based material news trend reporting system to quickly grasp the latest global trends in the materials field and establish effective business and research strategies. The automated process covers multilingual news collection and translation, ranking algorithm-based news recommendation, text summarization, generation of daily or weekly issue reports on material technology and industry by country, and email distribution.

The research team further expanded the system to a knowledge search and Q&A system that integrates the continuously accumulated latest material news and related documents. When users ask personalized material-related questions in natural language, they can receive the desired information and insights with over 90% accuracy in search and answers. More than 500 employees from about 10 affiliates within POSCO Group, including POSCO Holdings, use the system daily.

POSCO Holdings operates both the generative AI-based material news trend reporting and knowledge search Q&A systems using Google Cloud, selecting Gemini 1.5 Pro as the large language model (LLM) engine for generative AI.

In collaboration with POSCO, POSCO Holdings recently expanded the knowledge search Q&A system related to facility electrical knowledge. First, they processed over 30,000 pages of facility electrical documents accumulated within POSCO using multimodal AI specialized for POSCO documents. Based on this, a knowledge vector database was also constructed.

Subsequently, a system combining the in-house developed high-performance Retrieval-Augmented Generation (RAG) technology and the Gemini model was built. Users are evaluated to be able to easily and effectively obtain information search and answers to facility-related questions needed for work with about 90% accuracy.

Additionally, POSCO Holdings built the overall data pipeline of the in-house Applied AI platform, configured as a multi-cloud, on Google Cloud. POSCO Holdings is also conducting multimodal generative AI and AI-based simulation R&D to accelerate material development.

Im Woosang, Research Fellow (Executive Director) at POSCO Holdings Future Technology Research Institute, stated, "To properly utilize our own AI models, it was essential to build infrastructure capable of effectively serving data and models. Thanks to Google Cloud’s outstanding AI capabilities, rich experience, and close support from engineers, we were able to quickly and efficiently establish a foundation for generative AI innovation."

Jigi Sung, President of Google Cloud Korea, emphasized, "Collaboration with POSCO Group, which is leading the future materials and energy sectors, is an example showing the increasing adoption of generative AI in the manufacturing industry. We are honored that Google Cloud’s powerful generative AI technology and AI-optimized infrastructure can support POSCO Group’s research and development and business innovation."