Kakao Enterprise said on the 15th that GS Retail introduced an AI search and recommendation system based on Kakao Cloud and cut expense for training search and recommendation models by an average of about 35% per month.
Kakao Enterprise said GS Retail not only saved expense through Kakao Cloud but was also able to use about five times more graphics processing unit (GPU) memory than before, enabling more experiments and strengthening its search and recommendation models. GS Retail operates GS Shop for home shopping, GS25 convenience stores, and GS The Fresh supermarkets. GS Shop runs its search and recommendation system based on AI and Machine Learning from data pipeline development to model serving and quality monitoring.
Since 2024, GS Retail has operated its search and recommendation system and built and used its analytics environment around Kakao Cloud's GPU instances and Kubeflow. Since May last year, it has applied semantic search using a text embedding model trained on Kakao Cloud. The company said GS Shop improved its search quality through a recommendation model that blends a collaborative filtering (CF) method—which identifies semantic similarity between words and sentences via the text embedding model and calculates similarity between users and products based on user behavior data—and a sequence model suited to the context.
Kakao Cloud said it supports configuring GPU resources at the minimum unit, allowing GS Retail to scale resources flexibly as needed. It also said it helped execute the budget planned at the start of the year efficiently and stably by minimizing external uncertainties such as exchange rate fluctuations.
Kim Yo-han, head of the GS Retail AX Division for Home Shopping AX (executive director), said, "We were able to transfer and use the existing Kubeflow-based pipeline used in the search and recommendation system to Kakao Cloud as is, and we were very satisfied with its flexibility and scalability."
Lee Yong-min, head of Kakao Enterprise's cloud institutional sector, said, "As a cloud strategy partner that offers granular options tailored to customers' needs, we will do our best to build an AI infrastructure environment where customers can focus on their core competitiveness."