Data science operations startup Domino Data Lab Inc. today unveiled new capabilities in its popular Domino Cloud platform aimed at boosting the scalability of artificial intelligence applications and agents and make them more cost-effective.

Domino Cloud is a software-as-a-service platform that’s used by developers from companies such as Bayer AG and Moody’s Corp. to prepare datasets for machine learning projects and automate tasks such as deploying AI models and monitoring their performance.

The platform also provides comprehensive tools for AI governance, ensuring that neural networks work reliably and don’t expose sensitive data to unnecessary risks. During model development, it analyzes them for potential data breaches and other risks. Another key component is Domino Apps, which provide interactive tools for users to share and run AI applications and agents without extensive knowledge of DevOps practices.

Domino Apps is the focus of today’s update, which recognizes that many enterprises struggle to create value from their AI deployments due to an inability to scale them to support real-world use cases. It’s also designed to address “uncontrolled infrastructure spend,” which stems from the underutilization of compute resources, maintenance and excessive data movement, the company said.

To get around these problems, Domino said it’s introducing a new autoscaling capability in Domino Apps, plus a new compute grid tool to help customers get a handle on the spiraling production costs of AI.

The new autoscaling tools are meant to make AI more easily consumable for business users and will ensure that every new app and AI agent scales seamlessly. The platform will automatically dedicate more compute and storage resources as each one’s user base grows, without incurring unnecessary costs from underutilization.

In addition, there’s a new application discovery tool in Domino Apps, which can help workers to find existing apps and agents that may be relevant to their workflows. The idea is to facilitate greater consumption and help organizations squeeze more value out of their AI investments, the company said.

To ensure companies can keep a lid on their AI infrastructure costs, Domino is adding support for spot instances, enabling users to access compute resources for AI development, training and inference on demand, with cost savings of up to 60%. Spot instances are incredibly useful, because while they tend to be a bit more expensive, most cloud providers try to encourage companies to pay upfront to reserve instances, which leads to those resources being underutilized.

To aid further in cost-cutting, Domino said it’s now managing all customer data via data planes in the Domino Cloud platform. Its data plane technology serves as the execution layer for running and scaling AI and allows each single-tenant Domino Cloud environment to run securely across multiple geographic regions. As a result, customers can bring their compute resources closer to the data that powers their AI applications and agents, reducing data movement costs while simultaneously addressing any data sovereignty requirements they need to abide by.

Constellation Research Inc. analyst Holger Mueller said Domino seems to have gotten its priorities right. Today’s updates, he said, recognize that whoever can help enterprises to build AI applications faster and run them more cost-effectively is likely going to be successful. “It’s good to see a startup like Domino making inroads here, giving enterprises the ability to auto-scale their AI applications and use spot instances to run ad hoc workloads,” he said. “This is just the kind of thing enterprises want and need to accelerate AI adoption.”

Domino co-founder and Chief Executive Nick Elprin said the updates address key pain points for customers, who have expressed that while “AI experimentation is easy, finding value is much harder.”

The company said the new capabilities are rolling out for Domino Cloud on Amazon Web Services first of all, with support for additional cloud platforms, such as Google Cloud and Microsoft Azure, to come in upcoming releases.

Image: SiliconANGLE/Meta AI