Asking the right questions is at the core of Opinary – not only in our mission, but also within our team. Recently, our team members have questioned the status quo of our machine learning systems and therefore were able to improve it significantly.

Our two data scientists Doreen Sacker and ex-Opinarian Hector Otero Mediero, empowered by Torben Brodt, Pavel, Ario and Gleb, re-architected our recommendation engine. Our recommendation engine identifies which poll fits into an article on a publisher site, screening millions of polls and articles. With great hands-on help of people from Google and endocode, they migrated the system and made it faster, more cost efficient.

We migrated our machine learning pipeline from Google Kubernetes Engine (GKE) to Cloud Run. After making a few architectural changes, we reduced the time to generate a recommendation from 20 seconds to below one second, and realized a remarkable 50% cost reduction.

Read more about our journey and learnings on the Google Blog.