Getty Images/iStockphoto

Knime updates Business Hub to ease data science deployment

The vendor is adding a tool that enables continuous integration and deployment of AI and ML models to help organizations better operationalize data science efforts.

Knime has launched an extension of its Business Hub designed to make it easier and more secure for organizations to deploy and update data science products.

The extension, named Continuous Deployment for Data Science (CDDS), was unveiled on April 18 during Knime's Spring Summit in Berlin.

Now generally available as a set of Knime workflows and applications, CDDS enables organizations to automate the continuous integration and deployment (CI/CD) of machine learning models and other data science assets.

Enabling data science

Knime, which initially introduced its Business Hub in November 2022, is an open source analytics vendor based in Zurich, Switzerland, that also enables data science.

The Business Hub is a cloud-based environment where organizations can put security and data governance measures in place, share and collaborate on data science and analytics projects, and move those completed projects into production.

The platform essentially replaced Knime Server, which included many of the same capabilities but was not cloud-native and was therefore limited in its reach. While Server enabled sharing and collaboration within a department or other defined group of users, Business Hub enables broader sharing and collaboration across an entire organization.

Sharing and collaboration, meanwhile, are particularly vital to data science, which suffers from a high rate of failure due to factors including the lack of collaboration.

At the time Business Hub was launched, Knime said adding CI/CD capabilities for machine learning and augmented intelligence was part of the platform's roadmap.

Now, with the addition of CDDS, Business Hub enables simple deployment of data science projects and the automated monitoring and retraining of projects as needed once in production. In addition, it lets administrators oversee the entire deployment flow to ensure that only properly validated projects are put into production.

Donald Farmer, founder and analyst at TreeHive Strategy, noted that organizations often struggle to keep data science models current with up-to-date data. CDDS, therefore, is a valuable addition to Business Hub given its CI/CD automation capabilities.

"There is some value here," Farmer said. "The need to continuously train and update models is well-understood in theory -- data is constantly changing, after all. But knowing when and how to update a model efficiently can be tricky. It's like trying to tune an engine while the plane is flying. An automated and continuous process makes this more efficient."

The tool has the potential to be particularly useful for business users dabbling in data science, he continued.

"It ensures that models are updated in a timely manner when changes are needed," Farmer said. "In this way, important inflections in the accuracy of the model are not missed. This matters particularly to Knime with the rebranding to … Business Hub. If you are targeting business scenarios rather than data science research, then the models need reliable management and administration."

A sample screenshot from Knime.
A sample screenshot from Knime displays the analytics vendor's UI.

Similarly, Mike Leone, an analyst at TechTarget's Enterprise Strategy Group, said CDDS has the potential to help organizations address some of the problems they face when attempting to operationalize data science models.

In particular, he noted that it addresses multiple phases of the model lifecycle, including not only deployment but also improving models with fresh data.

"It adds robust capabilities to address the gaps organizations regularly face when starting to operationalize machine learning," Leone said. "So many organizations struggle with the last mile of AI, moving a model into production. But that's not the end of the journey. CDDS helps with the cyclical process of continuous deployment and improvement of models."

That includes ensuring ongoing governance is enforced, models can quickly react to data drift and models can be retrained models based on new data, he added.

While Knime's Business Hub was new late last year and the vendor is now adding new capabilities, the concept of a hub for data science and analytics projects is not unique to Knime.

For example, Google launched Analytics Hub in September 2022. ThoughtSpot, Tableau, Qlik and Microsoft Power BI are among other platforms that feature collaboration capabilities.

However, given Knime's emphasis on data science, whether first to bring collaboration capabilities to its platform or not, such capabilities are critical to the vendor's users.

And the addition of CDDS is a logical first extension of Business Hub following its initial release, according to Leone.

"It's adding capabilities that organizations are starting to ask for more and more," he said. "This really plays in the MLOps space where organizations aren't deploying one model and forgetting about it. They're deploying hundreds of models and need defined processes in place to iterate faster and with more confidence once the models are deployed."

More new capabilities

Beyond the launch of CDDS, Knime also unveiled new features for the Knime Analytics Platform, the vendor's free suite of business intelligence tools.

Knowing when and how to update a model efficiently can be tricky. It's like trying to tune an engine while the plane is flying. An automated and continuous process makes this more efficient.
Donald FarmerFounder and analyst, TreeHive Strategy

Version 5.0 of the platform includes a series of features designed to make it easier for new users to get started building analytics workflows.

For example, a tool called Starter Perspective aims to simplify the transition from spreadsheets to visual analysis with a curated set of nodes commonly used for data manipulation tasks such as data cleaning, merging and filtering. Another feature enables new users to explore 12 data wrangling examples on the Knime Community Hub.

While the update targets new users, Farmer noted that the performance of Knime's open source tools could be improved.

"The only substantial requests I hear from users are for better performance," he said. "I think the open source extensions are particularly slow."

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.

Dig Deeper on Data science and analytics

Data Management
Content Management