Cyber-resilience supplier Rubrik is buying AI agent development startup Predibase to accelerate agentic AI adoption.
Predibase is a 2021 startup founded by Google and Uber alumni, originally focused on providing machine learning (ML) tools to customers. It pivoted to GenAI LLM-based agent tools when ChatGPT arrived and its focus now is building tools to help customers move agents from pilot projects to at-scale production use. Predibase believes the future of generative AI is smaller, faster, and cheaper open source LLMs fine-tuned for specific tasks. The thinking is that competitive value in AI wouldn’t accrue to general-purpose APIs, but to how models are customized with proprietary data.
Rubrik co-founder, chairman, and CEO Bipul Sinha stated: “What the Predibase team has achieved with model training and serving infrastructure in the last few years is nothing short of remarkable. AI engineers and developers across the industry trust their expertise. Together, Rubrik and Predibase will drive agentic AI adoption around the world and unlock immediate value for our customers.”
The acquisition cost has not been revealed. A CNBC report suggests Rubrik paid between $100 million and $500 million for Predibase. In its Q3 2025 results, Rubrik reported it had cash, cash equivalents, and short-term investments of $632 million. It may be buying Predibase for cash – it could afford it as a mix of cash and shares, or just shares.
Rubrik has more than 6,000 enterprise customers. The great thing for backup suppliers providing proprietary data to AI models and agents is that they represent a single source of data, backed up as it is from multiple sources within a customer’s data and application estate, both on-prem and in the public cloud. If a customer wanted to provide this data to AI models and agents from the original, real-time sources, it would have to build AI pipelines to collect, filter, select, and move this data to a central place. Of course, it may already be doing this with some of it by utilizing data warehouses, lakes, and lakehouses. The backup folks, like Commvault, Cohesity, Rubrik, and others, provide an existing AI model and agent data collection point, like a data warehouse/lake/lakehouse but without the in-built analytics.
Now Rubrik, calling its backed-up data a data lake, wants to provide an agentic AI production pipeline, using Predibase tooling to take pilot projects to widespread production use. That software enables AI agent developers to fine-tune and deploy AI agents at scale and post-train them in the inferencing period.
The Predibase software has a proprietary post-training stack for customizing models with a highly optimized inference engine. There is a turbo serving engine and LoRAX (LoRA eXchange), an open source system for deploying personalized models at scale. It “allows users to pack hundreds of fine-tuned models into a single GPU and thus dramatically reduce the cost of serving.”
Sinha blogs: “Rubrik’s data lake provides the necessary access controls, permissions, and policies to securely power AI applications. Predibase delivers operational value on top of improved, customized models, enabling enterprises to build, pilot, and deploy AI safely at scale. When teams are equipped with Predibase and our secure data lake, they are empowered to accelerate AI adoption.”
Predibase CEO Devvret Rishi said: “We created Predibase to lift the barriers between an idea and production-ready AI. Today, many organizations still face challenges moving beyond the proof-of-concept stage. Predibase removes the hardest part of that journey and accelerates production-ready AI by giving teams an easy-to-use platform to tune models to their own data and run on an optimized inference stack. This unlocks more accurate results and faster models, all at lower cost.”
Rubrik noted that integrating Predibase will expand the work to secure and deploy GenAI applications that Rubrik is doing today with Amazon Bedrock, Azure OpenAI, and Google Agentspace.
Read more in a Predibase blog, where Rishi says Rubrik and Predibase will be “building towards a shared vision at the intersection of AI, security, and data,” and a Rubrik blog.
Rubrik has now added a third string to its bow, first providing backup, then security, and now agentic AI development tooling.
Bootnote
San Francisco-based Predibase was founded by original CEO and now chief scientific officer Piero Molino, CTO Travis Addair, and original chief product officer and now CEO Devvret Rishi.
Molino spent three and a half years at Uber as a natural language processing scientist and Addair worked at Uber for four years as a senior software engineer and tech lead manager. Both worked in machine learning projects there, such as Ludwig (auto-ML platform) and Michelangelo (ML-as-a-Service). Rishi was a Google product manager for Cloud AI during his time in machine learning.
Predibase raised an undisclosed seed round and then a $16.25 million A-round in May 2022, led by Greylock, with participation from Factory and angel investors, including Anthony Goldbloom. This was followed by an A expansion round in May 2023 for $12.2 million led by Felicis Ventures. Total known funding is $28.45 million and it helped take Predibase from beta product status to general availability. Customers include Checkr, Convirza, Marsh McLennan, Nubank, and many others.
We understand Predibase staff will be joining Rubrik.
Separately, startup Typedef is also developing software to take AI agent pilots to production deployment.
