Co-founder of Recombee (recommendation-as-a-service) and associate professor at CTU Prague where I run Recombeelab — a research group that publishes on recommender systems with access to real production data at scale. Building things at the intersection of ML research and production systems: beeFormer (transformer architecture for sequential recommendations), Repsys (open-source visual analytics for RecSys explainability), and p-book (adaptive interactive books where the navigation mode itself is personalized, not just the content order). Most interested in: the gap between what RecSys benchmarks measure and what actually matters in production, cold-start at scale, and making recommendation systems legible to non-experts. Happy to talk about anything in this space.