Google's new Vertex AI

At Google I/O 21, Google Cloud announced a new service, Vertex AI. In this post, we break down what Vertex AI is and link it back to some use cases for businesses.

What is Vertex AI?

Vertex AI is a managed machine learning platform that allows users to create, deploy and maintain their AI models.

As with most Cloud offerings, when you hear ‘managed’, this means that the cloud provider will provision the underlying hardware for you. This frees the end-user from having to create a Compute instance and install packages such as scikit learn or PyTorch.

Vertex AI aims to be an end-to-end platform for machine learning. The scope of Vertex AI is described in the following image.

Image showing that Vertex AI spans the whole machine learning lifecycle.

Navigating Vertex AI, it’s clear that there are two paths:

  • AutoML
  • Custom

AutoML gives users a couple of entry points into machine learning, depending on skill-level. Overall, with a drag-and-drop interface it can make machine learning incredibly accessible – which is great for smaller businesses.

With the option of writing custom code but still working within the end-to-end solution of Vertex AI benefiting those more experienced organisations who already have machine learning solutions or more advanced needs that AutoML doesn’t cater for.

The old re-packaged?

Google has long had Google AI Platform which includes Notebooks, Pipelines and Models. Vertex AI is more an amalgamation of these services with the addition of Datasets and Endpoints.

If you’re familiar with Amazon’s Sagemaker, Vertex AI is very similar to that – in fact, Craig Wiley, the director of product management for Google Cloud’s AI Platform came from Amazon where he was the General Manager for AWS Sagemaker.

The notable additions to Vertex AI are the Datasets and Prediction.

Vertex AI datasets allow you to create and manage datasets for a wide-array of machine learning tasks, such as classification, object detection, etc. This makes it easier to compare performance of various models.

Vertex Prediction helps with the ‘last mile’ of machine learning – the deployment. Vertex Prediction allows you to deploy your models for online serving through HTTP or batch prediction. Again, this is a managed service so there is no need to manage the underlying instances.

Who does this benefit?

The obvious answer would be anyone who already uses Google Cloud. But the more detailed answer is that I see benefits for a number of users:

  • Small businesses who want to benefit from machine learning, but don’t have the budget or ability to build custom models
  • Larger organisations who already do machine learning and want a managed way to train, compare and deploy machine learning models.

Inqyr Consulting works closely with Google Cloud and are happy to help you to see if Vertex AI is a good fit for your needs or help with your data solutions.