Skip navigation
machine learning banner next to machine learning logo Getty Images

Domino Data Lab Enhances MLOps Platform with Autoscaling Clusters, More

Domino Data Lab advances its namesake enterprise MLOps platform with advanced capabilities to help data scientists more effectively operationalize machine learning models at scale.

Domino Data Lab announced on Jan. 26 its new Domino 5.0 enterprise MLOps platform, providing organizations with enhanced capabilities for machine learning operations.

Based in San Francisco, Domino Data Lab was founded in 2013 with the goal of helping data scientists operationalize machine learning workloads. Included in the Domino 5.0 platform are new features designed to make MLOps easier to manage, including autoscaling clusters, data connectors and integrated monitoring functionality.

"Domino is a platform that covers the end-to-end lifecycle from exploration all the way to deployment and monitoring," Nick Elprin, CEO and co-founder of Domino Data Lab, told ITPro Today. "When we say enterprise, what we're really focused on is enabling large data science organizations who are doing data science at scale."

Domino 5.0 MLOps Platform Enables Autoscaling Clusters with Kubernetes

Domino enables data scientists to use whatever machine learning libraries, developer environments and frameworks they want, according to Elprin. The Domino MLOps platform helps scale the operations, and brings governance and reproducibility for machine learning workloads.

For workload orchestration, Elprin said Domino makes use of the Kubernetes container orchestration platform. As users start workloads, they get to pick their hardware, CPU, memory and GPU resources they want, while Domino orchestrates the workloads across a cluster, he explained.

In Domino 5.0, users will now benefit from an autoscaling capability. Elprin said autoscaling for Domino's distributed compute workloads is based on Kubernetes Horizontal Pod Autoscaler.

Domino Data LabDomino-5.jpg

Data Connectors Ease Enterprise MLOps Usability in Domino 5.0

Domino 5.0 also introduces a new set of data connectors that provide a mechanism to make it easier for data scientists to access and find data from different sources.

"What we've done with data connectors is give data science teams a way to define reusable components that describe how to connect to a data source," Elprin said.

The data services supported through the streamlined data connectors for the Domino 5.0 release are Snowflake, AWS Redshift and AWS S3. Domino plans to grow its list of data connectors in upcoming releases.

Integrated ML Workflow Monitoring Lands in Domino 5.0

Domino 5.0 also integrates monitoring capabilities for machine learning.

When users deploy a machine learning model with Domino 5.0, the platform can automatically monitor the model to look for potential issues, including performance or model drift, according to Elprin. He explained that model drift can occur when assumptions or conditions employed in a given model become inaccurate over time.

The Domino 5.0 enterprise MLOps platform automatically builds data capture pipelines to capture the live predictions that a machine learning model is scoring, he added. Domino automatically runs statistical analyses between the live prediction and the training data to look for potential deviations and drift.

"If our automated drift checks detect drift in your model, Domino 5.0 provides a button that data scientists can press to spin up a dev environment with the original state of their model when they deployed it and a data set of all their live prediction scoring data so they can quickly do an investigation into what caused that drift," Elprin said.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.