Rubrik to Purchase Predibase to Power Generative AI Growth

Rubrik plans to buy a startup led by a former Google Cloud AI product manager to address roadblocks in enterprise adoption of agentic AI technologies.

See Also: Taming Cryptographic Sprawl in a Post-Quantum World

The Silicon Valley-based data protection vendor said its proposed acquisition of San Francisco-based Predibase will help enterprises transition from proof-of-concept AI projects to scalable, production-ready applications that maintain trust, security and efficiency, according to Chief Business Officer Mike Tornincasa. He said the move will enhance Rubrik’s cyber resilience platform with cutting-edge AI skills.

“We see AI – generative agentic – as a business imperative, getting productivity from these capabilities,” he said. “There are two things really holding enterprise adoption back. One, governed, trusted data, which Rubrik has. And two, accuracy and optimization around cost and deployment. If I could venture a third, it would be a talent gap. Rubrik solves one, but two and three, we hope to solve with Predibase.”

Predibase, founded in 2021, employs 25 people and has raised $28.5 million of outside funding, having most recently completed a $12.2 million Series A expansion led by Felicis. The company has been led since October 2023 by Devvret Rishi, who served as Predibase’s chief product officer from the company’s inception in March 2021 until his promotion to the chief executive position (see: Identity Risk Remains Unclear Without Data Context).

How Rubrik, Predibase Can Make AI Deployments Safe, Affordable

Rubrik had identified generative and agentic AI as critical for enterprise productivity and scalability, but recognized persistent obstacles hindering AI adoption, specifically around data governance, deployment optimization and talent. While Rubrik’s Security Cloud addressed the need for trusted enterprise data, the company lacked a native solution like Predibase to improve model performance and cost-efficiency, he said.

“Ever since we started the company in 2021, we’ve had one fundamental belief, which was, ‘We want to democratize access toward AI,'” Rishi said. “And we thought that the most compelling AI applications are going to be brought by bringing models towards customer data. So our goal as a company is to help organizations graduate from POCs and pilots to production generative AI applications.”

Organizations often launch Gen AI pilot projects targeting internal use cases, where data access is easier to manage and security risks are lower. But expanding to production-level, especially external-facing use cases, remains elusive for many. Key barriers include difficulties in securely accessing and managing proprietary datasets, ensuring governance compliance and the accuracy of generative models.

“What we noticed, especially from some of our most tech-forward customers, is that a lot of generative AI applications get unlocked, not just on internal use cases, but also on external use cases,” Rishi told Information Security Media Group. “And there is a gap that needs to be bridged before these organizations are going to be comfortable doing that at scale for a lot of enterprises.”

At the core of this integration is Rubrik’s data lake, which combines enterprise data and metadata under strict compliance and governance, ensuring AI models access appropriate, high-fidelity data, Tornincasa said. Predibase adds a model-hosting and tuning layer, which Tornincasa said enables customers to customize and deploy models based on their proprietary data without leaving the secured environment.

“We believe we can bring an end-to-end solution to market that will be very compelling, not just for the Rubrik customer base, but for the market at large,” Tornincasa told ISMG.

Why Organizations Struggle to Scale AI Initiatives

Enterprises often struggle to go beyond the 70% to 80% accuracy threshold in generative AI applications, which Rishi said is insufficient for most production-grade use cases. Predibase provides a post-training stack that allows enterprises to bring their own data and fine-tune models to meet accuracy thresholds above 90%, which is considered the benchmark for trust and reliability for large-scale deployments.

“We build a stack that allows them to be able to bring data and examples of what they want their model to be able to do, and then combine that with the models that we host and deploy on Predibase and put those models directly on top of that data,” Rishi said.

Enterprises are wary of adopting AI at scale unless they can be sure that the data being used is accurate, permissioned and appropriately governed. Tornincasa said Rubrik brings a robust identity-based access management system that dictates who can use what data under what circumstances, preventing accidental or malicious misuse. This architecture will serve as the foundation for secure AI applications.

“Rubrik has a great understanding – as a need for building their cyber resilience product – of who has access toward which aspects of data,” Tornincasa said. “That’s going to be a foundational unlock for enterprises building generative AI applications. Rubrik also has a trusted customer base in the enterprise that are starting to come online with a lot of these generative AI applications.”

While technology-forward companies like those Predibase serves have the in-house expertise to deploy advanced AI models, Tornincasa said most enterprises do not. By integrating Predibase’s platform and specialized team, Rubrik will enable even less-technically sophisticated enterprises to benefit from AI. The goal is to reduce the learning curve and operational overhead, offering a turnkey AI tool, he said.

“If you look at where Predibase has had success, it’s primarily with the most innovative companies in the world, the fastest to adopt AI, those that have machine learning infrastructure expertise, those that have data scientists,” Tornincasa said. “But if you look at the rest of the world, they’re not as fortunate talent-wise or well positioned talent-wise to take advantage of these new capabilities.”