Seldon has been around for a while and has pivoted from being just a prediction platform to a machine learning (as a service) platform that can run multiple models and be used in many organisations. Though focussing on the Fintech arena, Seldon core can be deployed in any industry (a ML solution that runs in a Kubernetes environment).
Now Seldon is developing Seldon Deploy which is a UI based deployment system allowing models to deployed an tested automatically, this could be on live infrastructure with a percentage of traffic sent to the new model, or completely separate infrastructure, it's all up to how the deployment is specified. There are also levels of authority, so models can be developed by specialists, then made available for deployment and authorised by the correct people or teams.
If you're looking to do machine learning, then Seldon could ease the implantation, allowing data scientists to concentrate on the models themselves.