Wildlife Studios is one of the 10 largest mobile gaming companies in the world, we reiceive several millions of request per second in our services and process 2TB of data per day. A lot of our software products relies on ML models. This talk reviews the challenges and tradeoff we faced figuring out a standarized way for creating production ready ML models, reutilizing our current infrastructure (Gitlab, Kubernetes, Flux, AWS ECS, etc) and with minimal limitations and changes in the way our data scientist work.