Powering Millions of Real-Time Rankings at GetYourGuide

Delivering millions of personalized rankings daily is no easy feat, especially when it comes to incorporating real-time user behavior and ensuring model performance at scale. 

In this session, Mihail Douhaniaris and Martin Jewell from GetYourGuide share their journey of building a cutting-edge ranking system that incorporates real-time user behavior and ensures model performance at scale using Tecton's feature platform and Arize's observability tools. They'll discuss their innovative "discounted ranking impressions" feature and share insights on tackling challenges such as diverse feature types, real-time pipelines, cost-efficient serving, A/B testing, and drift detection.

Key takeaways:

• Learn how GetYourGuide built a real-time, personalized ranking system that handles 30 million predictions daily while adapting to user preferences.
• Discover best practices for feature engineering, including leveraging diverse data sources, defining feature views, and efficiently iterating on features.
• Gain insights into how GetYourGuide’s team reduced technical debt and increased time-to-value for their data and ML pipelines.
• Understand how GetYourGuide ensures consistent performance and identifies growth opportunities in their ranking system.