QCon London 2026: Behind Booking.com's AI Evolution: The Unpolished Story
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
Jabez Eliezer Manuel, Senior Principal Engineer at Booking.com, shared the company's 20-year AI and data evolution at QCon London 2026. Starting from Perl/MySQL in 2005, Booking.com scaled to ~6800 database instances and two massive on-premise Hadoop clusters before sunsetting Hadoop after a seven-year migration. Their ML inference platform now runs 480+ models delivering 400 billion predictions per day at under 20ms latency. The talk covered three layers: Data Management (MySQL architecture, Hadoop migration strategy), Machine Learning Engineering (evolution from Perl to agentic systems), and Domain Intelligence (GenAI, Content Intelligence, Recommendations, and Ranking). A key insight was their use of interleaving alongside A/B testing to run more ranking experiments with less traffic, and the challenge of replacing a historically 'undefeatable' ranking formula due to infrastructure constraints.
