Peer review only works if the right experts see the right papers. We built a recommender that finds them automatically, across hundreds of journals.
The challenge
Editors across hundreds of journals must find suitable expert reviewers for every submission. Done by hand, that search is slow and inconsistent, and it becomes a real bottleneck at the scale of a large journal portfolio.
What we built
A learning-to-rank recommender, built on Elasticsearch indexing and Python feature generation, that matches each paper to suitable expert reviewers across 500+ journals, surfacing the strongest candidates for the editor rather than making them hunt.
Editors now find roughly one in three reviewers automatically through the system.
The results
The manual reviewer search shrinks: editors find a meaningful share of reviewers automatically, speeding up peer review across the journal portfolio without lowering the bar on expertise.
Our role
We delivered the machine-learning and data-engineering capabilities within a cross-functional product squad, from indexing and feature generation through to the ranking model itself.