Banksalad Develops 'dataserving' Platform for Data Analysts to Build APIs Directly

Banksalad runs hundreds of types of analyses daily through its S3, Airflow, and Spark-based data pipelines, using them for product features such as 'News Updates' and 'User Analysis Tags.' However, previously, whenever analytics data needed to be put into production, a different architecture had to be designed each time, and the contribution of server engineers was essential.

To solve this problem, Banksalad developed dataserving, a platform that turns analytics tables into production APIs. Data analysts can directly build APIs equipped with condition-combination filter functionality by defining specifications using protobuf, with only minimal help from server engineers. For the backend, they adopted Amazon DocumentDB, which provides large-scale performance and a flexible JSON document model.

The first customer of dataserving was 'Financial Shopping,' which lets users explore MyData-based financial products and benefits. Data analysts directly performed pipeline development and specification definition, launching the product without server engineers. Since then, various data products such as 'My Asset Ranking,' 'Investment Strategies of Experts,' and 'Insurance Diagnosis' have been built on top of this platform.

Currently, dataserving is being improved to go beyond batch data serving and accommodate complex requirements such as access control, real-time capability, and user event stream processing. Through this, it enhances the stability and operational efficiency of analytics data and supports the launch of an even wider variety of data products.