What is Google's AI Hub

AI Hub and Kubeflow Pipelines: Google wants to make AI components easier to use

"Our goal is to make AI accessible to all companies", writes Google on its cloud blog. That is why the company is working, among other things, to significantly reduce the entry barriers for using the company from its point of view and is now offering the results of this work as a service on its cloud platform. This includes, among other things, the new AI hub, which will serve Google as a central destination for plug-and-play content in the field of machine learning.

  1. L-Bank, Karlsruhe
  2. UKV - Union Krankenversicherung AG, Saarbrücken

In the AI ​​Hub, Google bundles a large number of different resources that are required for work in this area. This includes ready-made machine learning pipelines, Jupyter notebooks, Tensorflow modules and other offers. The finished pipelines, including algorithms, are based, among other things, on research work and the productive use of technology at Google itself, so that companies can quickly recycle and process them. In addition, this offer should enable companies to set up a kind of private and secure contact point for their own AI resources internally, where they can be uploaded and shared.

But Google writes: “It's not enough to provide a place where organizations can discover, share, and reuse ML resources. You also need a way to build and package those resources in a way that will be useful to most internal users ". This is what the new Kubeflow pipelines are supposed to offer.

This technology is based on the open source project Kubeflow, which in turn is based on Jupyter notebooks and the in-house machine learning framework Tensorflow. As mentioned, Kubeflow is supposed to make packaging ML code as easy as building other apps. The so-called pipelines are supposed to take over the orchestration of the various components. Likewise, experiments should be made easier than before and individual components can be reused and connected with one another more easily.