Google LLC is helping companies embrace the idea of what’s known as machine learning operations with the launch of a new managed platform today that it says will accelerate the deployment and maintenance of artificial intelligence models.

MLOps is to machine learning what DevOps is to application development. With MLOps, the idea is to add discipline to the development and deployment of machine learning models by defining processes that make machine learning development more reliable and productive.

The discipline brings together all of the engineering pieces that are required to deploy, run and train AI models. That includes steps such as data collection, data verification, feature engineering, resource management, configuration, model analysis and so on.

Google said today that its new Vertex AI platform will be key to enabling MLOps. Vertex AI is a managed machine learning platform that can train AI models using 80% fewer lines of code than alternative platforms, informed by the years of experience Google has in building, deploying and maintaining machine learning models for its own use.

The key advantage of Vertex AI is that it brings together all of the Google Cloud services necessary for building machine learning models under a single, unified user interface and application programming interface. Using this single environment to build, train and deploy the models at scale, companies will be able to move them from experimentation to production much faster, discover patterns and anomalies more easily, and make better predictions and decisions, Google said.

Andrew Moore, vice president and general manager of cloud AI and industry solutions at Google Cloud, said the company is trying to kick start an industrywide shift that will get everyone serious about moving AI from “pilot purgatory and into full-scale production.”

Google itself ran into that issue just a couple of years ago, Craig Wiley, director of product management for Google Cloud AI, said in an interview with SiliconANGLE. “We had a collection of exciting capabilities, but making them work together felt harder than the problems they were intended to solve.”

And if Google, well-known for its AI and machine learning chops, was having issues, enterprises were even worse off. “Machine learning in the enterprise was in crisis,” Wiley said. “Companies are not seeing the return on investment with AI.”

Vertex AI includes the same AI toolkit that Google uses internally, including computer vision, language, conversation and structured data. It also adds MLOps features to help with the creation of new machine learning models. They include Vertex Vizier, which is used to increase the rate of experimentation, the Vertex Feature Store for serving, sharing and reusing machine learning features, and Vertex Experiments to speed up model deployment through faster model selection.

The Vertex Continuous Monitoring and Vertex Pipelines features, meanwhile, help companies to maintain their machine learning models and streamline the entire ML workflow.

Vertex AI isn’t coming into an empty market. Amazon Web Services Inc. has offered its own SafeMaker for several years now.

Forrester Research Inc. analyst Kjell Carlsson told SiliconANGLE that Google is offering the first integrated, end-to-end platform that covers not just the full AI model development and operationalization lifecycle, but the data engineering side too.

“It’s a unified AI platform that lowers the barrier to entry for MLOps, or the creation and deployment of AI applications, and ultimately, AI-fueled digital transformation,” Carlsson said. “Before this, there were no good solutions out there that provided all the professional-data scientist-and-data engineer-grade capabilities in an integrated platform. Enterprises had to stitch together a stack of disjointed tools and technologies making it hard for nearly all enterprises to get the full value from their AI initiatives.”

Chirag Dekate, an analyst with Gartner Inc., told SiliconANGLE that research has shown that most companies struggle to get AI into production.

“Nearly one in two pilots never make it into production and of the ones that do, it takes nearly eight to nine months on average,” Dekate said. “Gartner’s view is that enterprises should build platforms that enable them to develop, manage, deploy and govern thousands of models in production.”

Vertex AI addresses all four steps, first by customizing and automating processes such as data ingest, data analysis, data transform, model training, validation and deployment, helping accelerate the way new AI models are nurtured, Dekate said. Vertex AI also makes it possible for enterprises to use Google’s best-in-class algorithms for management, while its AI Platforms and MLOps suite accelerates the move into production. Finally, he said, it incorporates a unified AI governance ecosystem on the compliance side.

Early adopters of Vertex AI have given the platform a big thumbs up too. ModiFace Inc., an augmented reality platform provider owned by the cosmetics giant L’Oréal S.A., enables consumers to test new beauty products virtually and in real time. It said it trained all of its AI models that enable this using the Vertex AI platform.

One example of ModiFace’s services is its skin diagnostic tool, an AI model trained on thousands of images of people’s skin that helps to create precise, tailor-made skincare routines for each of its customers. “With more and more of our users looking for information at home, on their phone, or at any other touchpoint, Vertex AI allowed us to create technology that is incredibly close to actually trying the product in real life,” said ModiFace Chief Operating Officer Jeff Houghton.

With reporting from Robert Hof

Image: vectorjuice/freepik