Oracle Corp. today launched a new cloud-based data science platform that it says provides a toolkit for analytics and artificial intelligence projects in the enterprise.

The Cloud Data Science Platform, announced today at a London event by Oracle Chief Executive Officer Safra Catz (pictured), comprises seven distinct services and feature sets. At the center of the suite is Cloud Infrastructure Data Science. It’s a virtual workbench in which engineers can build machine learning models with assistance from an AI that automate parts of the workflow.

An automated work environment

The first task the product automates is choosing the right machine learning model for a project. Cloud Infrastructure Data Science can take multiple algorithms, as well as multiple configurations of each of those algorithms, and run tests to find the most suitable candidate. It also automates feature engineering, the process of determining which data points an AI should consider most seriously when making decisions.

An accompanying model assessment tool enables engineers to check that their neural network lives up to expectations. According to Oracle, it generates visualizations that show how well a model  is processing the data being thrown at it. Engineers can also follow how performance changes in production to identify potentially unwanted fluctuations.

Automating the AI development workflow is one of two main priorities Oracle has set for the solution. The other is “adding strong team support for collaboration to help ensure that data science projects deliver real value to businesses,” according to Greg Pavlik, the head of product development for the company’s Data and AI Services group.

Oracle has to this end equipped the solution with a host of collaboration features. There’s a dashboard that explains how much weight a model attributes to each data point it evaluates in a decision, which helps engineers provide transparency into their software for stakeholders. Also included: a catalog through which team members can exchange models, shared project folders and security controls for managing who is allowed to access what.

More AI features

The Cloud Data Science Platform of which the Cloud Data Science Platform is part is launching with several other components. Among them is a set of AI features for Oracle’s flagship product, the Autonomous Database. 

A capability called OML4Py will let enterprises run machine learning models directly on the system so data can be processed locally instead having to be moved to a separate, dedicated AI environment. This should save bandwidth while freeing up some time for administrators. A second feature called OML4Py AutoML automatically recommends the best AI model for the specific information a company is processing.

AI is playing an increasingly strategic role in Oracle’s product roadmap. The company has equipped the Autonomous Database with AI features for automating day-to-day administration, a move credited with helping to buoy customer interest in the system even as Oracle faces rising competition from bigger cloud players.

Oracle also unveiled several new analytics services as part of today’s announcement. There’s Data Flow, a managed version of Spark, Big Data Service, a cloud-based implementation of Cloudera Inc.’s Hadoop implementation, and a service called Cloud SQL that allows analysts to run queries across multiple systems of record at once.

Photo: Oracle/Flickr