Google LLC today announced the general availability of AI Platform Prediction, a service that allows companies to host machine learning models on its public cloud without having to worry about infrastructure management.

Setting up a production-grade machine learning environment manually can be technically challenging even for large enterprises. With AI Platform Prediction, Google promises to eliminate most of the hassle.

The offering allows customers to create a machine learning environment atop Google’s GKE managed Kubernetes service without having to set up and maintain the deployment on their own. AI Platform Prediction is also receiving several new features on occasion of its release into general availability. 

To improve security, Google has added the ability to create a so-called perimeter around machine learning models to isolate them from the rest of a company’s cloud environment. Administrators can configure the perimeter to only let a model interact with resources and workloads it strictly needs to access. This kind of isolation comes handy in the event of a breach because it makes it harder for hackers to move deeper into the corporate network by hopping from one application to the next.

Another new AI Platform Prediction feature is Resource Metrics. In Google’s Cloud Console and Stackdriver monitoring tools, administrators can now view models’ cloud infrastructure utilization to identify opportunities to lower hardware costs or optimize performance.

Different types of artificial intelligence initiatives often require different development frameworks. To expand the range of projects companies can run on AI Platform Prediction, Google is adding improved support for models created with the XGBoost and scikit frameworks. XGBoost is used to build models based on a method called gradient boosting, which is an alternative to deep learning useful for analyzing numerical data like spreadsheets. Scikit is a relatively simple AI creation tool that prioritizes ease of use. 

“AI Platform makes it simple to deploy models trained using these frameworks with just a few clicks — we’ll handle the complexity of the serving infrastructure on the hardware of your choice,” Google engineers Bhupesh Chandra and Robbie Haertel wrote in a blog post. 

Alongside the machine learning enhancements, Google today debuted a set of more general-purpose monitoring features for tracking the health of cloud instances. The company offers software agents that administrators can install on their instances to track metrics such as memory usage. Administrators are now receiving access to an improved monitoring agent for Windows-based virtual machines, as well as a time-saving tool that allows agents to be installed and updated in bulk.

“With as little as one command, you can create a policy that governs existing and new VMs, ensuring proper installation and optional auto-upgrade of both agents,” explained Google Cloud product manager Morgan McLean. 

Image: Google