Google Cloud platform recently announced the launch of a new enterprise-grade service, Cloud AI Platform Pipelines that will facilitate developers with a single tool to deploy their ML pipelines, along with tools for monitoring and auditing them. The new offering will simplify the deployment of complicated ML workflows that generally have several moving and interdependent parts.

In an announcement, Google stated that when a user is just prototyping a machine learning model in a notebook, it may seem quite straightforward, but if one starts focusing on the other pieces required to make a machine learning workflow sustainable and scalable, that is where things tend to become more complex. And as the complexity increases, building a repeatable and auditable process becomes even harder.

To simplify the complexity, pipelines can be brought to work which offers developers the ability to build these repeatable processes. Google further illustrated in the statement that there are two parts to the service: the tools for building and debugging pipelines and an infrastructure for deploying and running workflows.

The service aids in manually configuring Kubeflow Pipelines and automates the processes such as setting up Google Kubernetes Engine clusters and storage. It also leverages TensorFlow Extended to structure TensorFlow-based workflows and the Argo workflow engine for running the pipelines.

In addition to the services provides by the infrastructure, users will also facilitated with visual tools for versioning, artifact tracking, building the pipelines, and more.

Apparently, getting all this started only takes a few clicks, though actually configuring the pipelines is not easy. However, Google Cloud is adding a bit more of complexity, given that users can utilize both the TensorFlow Extended SDK and Kubeflow Pipelines SDK for authoring pipelines.

Source Credit: https://techcrunch.com/2020/03/11/google-cloud-launches-new-mlops-tools-for-deploying-ml-pipelines/