Equinor has many machines--many, many machines. And we would like to know if they are going to fail before they fail; otherwise, we lose a lot of money. Since we happen to have so many machines, we can not apply traditional data-science, which is focused on making an awesome model for a single problem. We need to build and deploy thousands of adequate models automatically. In order to do this we have created the open-source project Gordo with the goal of building and serving thousands of machine learning models. Nobody else uses it, but if you happen to have thousands of machines, millions of time-series, a big kubernetes budget, and the need for anomaly detection or predictions, maybe you should start using it?