Learning to Rank with Clojure – how we run our ML pipelines reliably (www.otto.de)

OTTO has announced its implementation of Learning to Rank using Clojure, moving away from traditional Gradient Boosted Decision Trees (GBDTs) to a Deep Neural Network that enhances product relevance in e-commerce search. This shift is significant for the AI/ML community as it highlights an unconventional choice of programming language—Clojure—known for its functional programming capabilities, concurrency, and resilience when handling complex machine learning pipelines. By leveraging Clojure’s robust support for modern software development practices, the company aims to deliver a more efficient and maintainable search experience that adapts dynamically based on user interactions. Clojure’s integration allows for the construction of data pipelines that process large datasets more effectively and with lower operational costs, thanks to its use of Protobuf for self-documenting, high-efficiency data handling. Additionally, the adoption of the Polylith architecture fosters clear dependency management through a monorepo approach, ensuring that all components remain updated and tested simultaneously. As OTTO embarks on future developments—like real-time personalized rankings—Clojure’s strengths appear poised to drive innovation in their machine learning efforts, suggesting exciting possibilities for the broader AI/ML landscape.