Pure Storage Pure Storage and Nvidia jointly announced a new RAG pipeline for optimizing large language models for the enterprise. The companies also revealed that Pure Storage has achieved Nvidia OVX certification. Optimized RAG Pipeline for AI Inference For generative AI to live up to its promise of transforming the enterprise, it first needs to meet the needs of the enterprise. Large language models need business-specific context to minimize hallucinations and deliver targeted business value. Retrieval-augmented generation, or RAG, is how LLMs are fine-tuned with business-specific data. At GTC, Pure Storage and Nvidia demonstrated an RAG pipeline that simplifies and accelerates the deployment of RAG for AI inference. The pipeline is designed to augment LLMs by integrating them with external, specific, and proprietary data sources, improving the models' accuracy, timeliness, and relevance. Utilizing Nvidia GPUs for compute tasks and Pure Storage’s FlashBlade//Sall-flash enterprise storage, the pipeline efficiently manages a large vector database and its associated raw data. This integration ensures that the data processing and retrieval tasks run swiftly and accurately. Pure conducted scalability tests for GenAI RAG embedding and indexing, comparing the performance of using local flash storage versus the Pure Storage FlashBlade//S with a native S3 interfaces. Pure’s tests showed that document embedding and indexing processes completed 36% more quickly when utilizing the FlashBlade//S, proving that fast, networked all-flash storage can significantly accelerate RAG document embedding activities. Taking things further, Pure Storage and Nvidia announced they are working together to deliver vertical-specific RAG solutions. Initially, they are making a financial services RAG solution available to provide quick and accurate summaries and analyses of large datasets, surpassing standard LLM performance. Plans to release similar solutions for healthcare and the public sector are underway. Certified Nvidia OVX Reference Architecture Nvidia’s OVX certification ensures that AI infrastructure, including servers and storage, can meet the specific performance, reliability, and compatibility standards required for operating on Nvidia’s advanced computing environments. While Nvidia OVX is a platform designed to support Omniverse applications—Nvidia's real-time simulation and collaboration platform for 3D workflows—the OVX certification indicates that the server hardware is optimized for these demanding workloads. Pure Storage's OVX certification, achieved in collaboration with Nvidia, signifies that Pure Storage's storage solutions, designed explicitly for OVX servers powered by Nvidia L40S GPUs and integrated with FlashBlade//S, have met rigorous Nvidia standards for performance, reliability, and compatibility. The certification is essential for enterprises making infrastructure choices for deploying AI applications, including generative AI and RAG for customizing LLMs. This certification complements Pure Storage's previous certification for Nvidia DGX BasePOD, providing additional options for AI customers. Analyst’s Take As network speeds increase, quickly moving data from storage to the GPU becomes critical. Simplifying enterprise AI tasks, like RAG, goes a long way toward improving the efficiency of the entire data pipeline. It's no surprise to see Nvidia working with technology providers like Pure to optimize the RAG pipeline. Nvidia is doing similar work with WEKA, NetApp, Dell Technologies, and DDN. Enhancing the performance of the RAG pipeline allows enterprises to process their entire document repositories into a vector database and query it in seconds, enabling faster time-to-value from their chosen AI applications. The flexibility and scalability of Pure Storage solutions make them ideal for businesses looking to incorporate LLM and RAG into their production processes. As the demand for AI continues to grow, enterprises will increasingly look for infrastructure to support AI models' rapid development and deployment. Pure Storage's OVX certification positions it as a go-to choice for businesses leveraging AI to gain a competitive advantage. Pure Storage’s innovations in the AI space demonstrate that the company is a proactive player in addressing the complex demands of AI-driven enterprises, with a clear strategy for sustained growth and market penetration. Disclosure: Steve McDowell is an industry analyst, and NAND Research is an industry analyst firm that engages in, or has engaged in, research, analysis and advisory services with many technology companies, including those mentioned in this article. Mr. McDowell does not hold any equity positions with any company mentioned in this article.