Kakao Ad Recommendation Team Operates MLOps-Based Serving System for Large-Scale RTB Optimization

The Kakao Ad Recommendation Team is central to Kakao's advertising business, responsible for developing optimization logic and serving systems using AD TECH. To maximize the performance of programmatic algorithms, the team applies advanced data science technologies to research and develop six optimization models (user responsiveness prediction, bidding strategy, optimization targeting, exploration, recommendation ranking, and abuse detection).

Additionally, to reliably provide real-time recommendation services based on these models, the team develops and operates a massive serving system. They build a Data Pipeline, model training and experimentation system, and metric monitoring system, operating the entire system organically within an MLOps structure. Data processing utilizes Kafka, Spark, and Hadoop, while modeling uses PyTorch, TensorFlow, and others. Regarding organizational culture, the team encourages remote and flexible work arrangements, aims for a horizontal culture, and supports members' technical growth through internal study groups.