In the ever-evolving landscape of technology, few companies have undergone as remarkable a transformation as Booking.com. The journey of this travel giant, from its humble beginnings in 2005 to its current state as a leader in AI-driven innovation, is a testament to the power of adaptability and forward-thinking. As Jabez Eliezer Manuel, Senior Principal Engineer at Booking.com, revealed in his presentation at QCon London 2026, the story of Booking.com's AI evolution is not just about technological advancements but also about the lessons learned and the challenges overcome.
One of the key insights from Manuel's presentation is the importance of a unified command center. In the early days, Booking.com's tech stack was built on Perl libraries and MySQL, which offered asynchronous replication and commercial support. However, as the company grew, it became clear that a more scalable and efficient system was needed. The introduction of Apache Hadoop in 2011 was a significant milestone, but it soon became apparent that Hadoop had its limitations. The cracks in the system, such as noisy neighbors and capacity issues, highlighted the need for a more robust and flexible solution.
The migration strategy was a meticulous process, involving five phases: mapping the entire ecosystem, analyzing usage to reduce scope, applying the Google Search PageRank algorithm, migrating in waves, and finally, phasing out Hadoop. This approach ensured a smooth transition and minimized disruption to the business. The key to success, according to Manuel, was a unified command center, which allowed for better coordination and control throughout the migration process.
The evolution of Booking.com's machine learning stack is another fascinating aspect of the story. From Perl libraries and MySQL in 2005 to agentic systems in 2025, the company has embraced a wide range of technologies along the way. Apache Oozie, Apache Spark, MLlib, H2O.ai, deep learning, and GenAI have all played a role in shaping the company's machine learning capabilities. The pivotal year of 2015 marked a turning point, as Booking.com solved two core problems: real-time predictions using online inference at scale and feature engineering for training and inference.
As of 2024, Booking.com's current machine learning inference platform boasts an impressive array of capabilities. With more than 480 machine learning models, 400 billion predictions per day, and a latency of less than 20 milliseconds, the company is at the forefront of AI-driven innovation. The use of domain-specific machine learning platforms, such as GenAI, Content Intelligence, and Recommendations, has enabled Booking.com to deliver personalized and efficient services to its customers.
However, the story of Booking.com's AI evolution is not without its challenges. The ranking formula, for example, was initially a simple function that included parameters such as bookings and the number of views. As the company attempted to replace this formula with machine learning, it discovered that the formula was 'undefeatable' due to infrastructure limitations. The need for a more sophisticated and flexible ranking system highlighted the complexity of AI-driven innovation and the importance of continuous learning and adaptation.
In conclusion, the story of Booking.com's AI evolution is a fascinating journey of technological innovation, adaptability, and forward-thinking. As the company continues to push the boundaries of AI-driven innovation, it serves as an inspiration to other businesses looking to embrace the power of AI. The lessons learned and the challenges overcome along the way provide valuable insights for anyone looking to navigate the complex and ever-evolving landscape of technology. From the importance of a unified command center to the need for continuous learning and adaptation, Booking.com's story is a testament to the power of innovation and the importance of staying ahead of the curve.
