Kim Young-jin (김영진), chief technology officer at NOL UNIVERSE, shared a strategy for making meaningful use of vast amounts of data and outlined the company’s future vision in a keynote speech at the “AI Days Seoul” conference hosted by cloud data platform company Databricks on Tuesday.

NOL UNIVERSE has accumulated large-scale data across lodging, aviation, leisure and entertainment based on its 13 million members. It is speeding up efforts to use that data meaningfully by using the Databricks platform.

Kim said NOL UNIVERSE’s leadership supported data-driven decision-making, but it lacked the ability to quickly and easily access needed data or effectively eliminate data silos.

A diagnostic review conducted with Databricks identified three tasks: a discovery issue that makes it hard to find needed data within vast datasets, a trust issue that makes it hard to be confident the data found can be relied on, and a complexity issue that prevents an integrated view across multiple services such as lodging, aviation and leisure. This left teams needing data to request it from specialists and wait for days.

To address these problems, NOL UNIVERSE built a centralized feature store called Lynx. It had four main goals.

Kim said the goals were that every member can generate and contribute data, all development is controlled through code reviews, a centralized hub is created for search, and the entire process from feature creation to deployment and management is automated. He described it as a way of using Databricks not as a simple analytics tool but as a data platform.

Kim said that before adopting the feature store, a data lake, a Presto-based query engine, an Airflow-based scheduler, a data hub-based catalog and Jupyter-based notebooks all operated separately. He said it was hard to connect systems and management was complex. He added that integrating these functions into Databricks significantly sped up development.

The feature store is being used in business in a meaningful way. Kim said a user segmentation platform using Lynx-based features has been applied to marketing and personalised services. He said targeting tasks that once required requesting help from data analysts and waiting for days can now be executed with a few clicks. Kim also emphasised building a global data platform spanning the entire Yanolja Group as part of its future data strategy.

He said the work involves connecting affiliate data scattered around the world, from Tel Aviv in Israel and India to the United States and Brazil. He said the company is currently pursuing a global data mesh implementation project using Databricks. He added it will use data and AI to create an environment where people can play 10 times more easily and enjoyably wherever they travel.