In a major product refresh, DataRobot Inc. today debuted a new edition of its platform for building artificial intelligence models that it says will enable enterprises to develop machine learning software faster.

The startup says that the platform, AI Cloud, can function as the central hub for a company’s AI initiatives. It provides features for all the key coding and software deployment tasks involved in building machine learning applications. Organizations can run AI Cloud in all the major public clouds and their on-premises data centers. 

The launch is DataRobot’s broadest product update since its most recent $300 million funding round earlier this year. The round, which was led by Altimeter Capital and Tiger Global, gave the startup a valuation of $6.3 billion. 

“AI will reshape every industry, every business service, every customer interaction,” stated DataRobot Chief Executive Officer Dan Wright. “DataRobot AI Cloud captures insights and learnings from years of technological innovation and partnering closely with our customers.”

DataRobot targets both data scientists and business users as part of its go-to-market strategy. For business users, DataRobot provides a point-and-click AI builder that makes it possible to create a machine learning model without any coding.

The tool evaluates the task that the user wishes to automate using AI, then automatically finds a suitable neural network from a library of pre-packaged algorithms provided by DataRobot. The selected neural network is sequentially trained on a similarly automated basis. 

AI Cloud introduces a feature called Composable ML that, according to DataRobot, will make its no-code AI builder useful not only for business users but also data scientists. Often, data scientists prefer coding their own custom neural networks instead of using an algorithm supplied by a development automation tool.

Composable ML allows them to blend the two approaches. Data scientists can generate a neural network using DataRobot’s no-code AI builder and customize it to their requirements with custom code, an approach that DataRobot promises is faster than building everything from scratch.

The startup says it’s addressing a major challenge in enterprise AI initiatives. According to research from DataRobot’s Algorithmia unit, 87% of companies cite long deployment times as a challenge in machine learning projects. Speeding the process by reducing manual work for data scientists can in theory help enterprises realize a return on their machine learning investments faster. 

The AI Cloud platform is aimed at “enabling users of all skill sets to deliver more models, faster,” Wright said. 

A particularly time-consuming aspect of AI development is data preparation. Before a company can send a set of business records to a machine learning model for analysis, it has to import the records from the system where they’re stored, filter any errors they contain and, in some cases, modify the way the information is structured. DataRobot is releasing a feature called DataRobot Pipelines as part of Cloud AI that allows customers to create software workflows for automating the process. 

Another set of new features is aimed at making the management of AI models easier once development is complete and they’re running in production. A capability called Bias Monitoring helps data scientists detect when bias occurs in an AI model. Another new addition,  Continuous AI, enables companies to automatically retrain their machine learning applications on a new dataset.

AI models are often used for the same types of business tasks across different companies. For example, two competing retailers might both have a machine learning application that predicts when product shipments from suppliers will arrive.

DataRobot is rolling out templates for common machine learning use cases to reduce the amount of work involved in building AI applications. The templates, which DataRobot calls Pathfinder Solution Accelerators, are aimed at further reducing the amount of time required to complete AI projects in the enterprise.

Photo: DataRobot