Rubrik acquires Predibase to help companies transition AI prototypes to production-ready applications. The acquisition combines data governance and optimized model training.
The acquisition is intended to accelerate agentic AI adoption by offering organizations a simpler path from experimental prototypes to scalable production environments. Predibase was founded by former AI engineers from Google and Uber. The platform specializes in fine-tuning open source models for production use. The technology combines a patented post-training stack with an optimized inference engine, enabling teams to support various use cases without incurring exponential infrastructure costs.
Breaking the proof-of-concept barrier
According to Gartner research, more than half of all AI projects never make it to production. On average, it takes approximately eight months to transition from an AI prototype to production. The combination of Predibase and Rubrik aims to address these challenges by reducing risks around data access, improving model accuracy, and lowering infrastructure costs.
“We created Predibase to lift the barriers between an idea and production-ready AI,” said Devvret Rishi, founder and CEO of Predibase. The platform promises up to 80 percent cost savings compared to hosting foundation models, while improving performance by more than two times.
Integration with existing cloud services
The acquisition aligns with Rubrik’s existing work on GenAI applications through Amazon Bedrock, Azure OpenAI, and Google Agentspace. By integrating Predibase, organizations gain access to optimized, fine-tuned models combined with Rubrik’s secure and regulated data lake.
The LoRA eXchange project for personalized models at scale is an important part of Predibase technology. This open source system makes it possible to support different users, use cases, and departments without increasing infrastructure complexity.
Tip: Rubrik AI assistant Ruby helps increase cyber resilience