Since May 18, Vertex AI from Google Cloud is available: this MLOps platform allows to deploy AI models as well as to maintain them. The tool aims to simplify the process of creating and deploying machine learning models. The Vertex AI announcement comes during the annual Google I/O conference where other new features such as the fourth generation of TPU, MUM technology or the LaMDAchatbot were presented.

A unified MLOps platform for Google’s Machine Learning services

Machine Learning Operations (MLOps) provide a framework for all functions designed to implement an AI enterprise strategy, in the same way that DevOps practices have revolutionized software development and its industrialization.

At Google I/O 2021, the company announced the availability of its MLOps platform, called Vertex AI. It presents, through an API and a web interface, all the services offered by Google Cloud, such as AI Plateform and AutoML, to train machine learning (ML) models. According to the group, the tool will simplify the creation and deployment of ML models at different scales and can be used without the need for training in machine learning.

Andrew Moore, vice president and general manager of Cloud AI and Industry Solutions at Google Cloud, said of the platform’s capabilities:

“We’re very proud of what we’re offering with this platform, as it enables serious deployments for a new generation of AI that will allow data scientists and engineers to do rewarding and creative work.”

A multitude of features and use cases

Vertex AI offers a wide range of features. Among those mentioned by Google are:

• Auto ML for custom ML model development.
• Vertex VizierVertex Edge Manager for automating hyper-parameter optimization.
• Vertex Feature StoreVertex Edge Manager for classifying, distributing, sharing and reusing machine learning model parameters
• Deep learning VM Images for the creation of virtual machines.
• Vertex Continuous Monitoring and Pipelinesfor the maintenance and repeatability of ML models. They streamline the machine learning workflow.
• Vertex Data LabelingIt also includes a new feature called “Label Quality” to improve the quality of labels and labeling and thus create better ML models.
• Vertex Edge Manager specifically dedicated to Machine Learning at the edge (IoT and Edge Computing).
• Vertex Tensorboard which uses Tensorboard to track experimentation metrics during the working phases of an ML model.

The image below summarizes, through a detailed diagram, all the new features and possibilities offered by Vertex AI:

VertexAI is available at: https: //cloud.google.com/vertex-ai

Translated from Google I/O : annonce de Vertex AI, la nouvelle plateforme MLOps de Google Cloud