© Reuters.
SANTA CLARA, Calif. – Pure Storage (NYSE: NYSE:), in partnership with NVIDIA (NASDAQ:), announced today the release of new validated reference architectures designed to enhance generative AI use cases. This development includes a NVIDIA OVX-ready validated reference architecture, which aims to provide enterprises with a framework for managing the data and compute requirements essential for AI deployments.
The collaboration between Pure Storage and NVIDIA has produced various technological advancements to cater to the increasing demand for AI applications within enterprises. The new validated designs and proofs of concept include a Retrieval-Augmented Generation (RAG) Pipeline for AI Inference. This pipeline utilizes NVIDIA NeMo Retriever microservices and GPUs in conjunction with Pure Storage’s all-flash enterprise storage to expedite the time to insight for AI training using internal data, thereby reducing the need for frequent retraining of large language models (LLMs).
Additionally, Pure Storage has achieved certification for its NVIDIA OVX Server Storage Reference Architecture, offering enterprise customers and partners a validated storage framework that is optimized for cost and performance in AI hardware and software solutions. This complements Pure’s previous certification for the NVIDIA DGX BasePOD.
The company is also focusing on developing industry-specific RAG solutions, beginning with a financial services RAG that aims to provide more accurate summarization and querying of massive datasets. Plans to release additional RAG solutions for healthcare and the public sector are underway.
To support AI deployments, Pure Storage is investing in its AI partner ecosystem, engaging with independent software vendors (ISVs) like Run.AI and Weights & Biases, and working with AI-focused reseller and service partners such as ePlus, Insight, and World Wide Technology (WWT).
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.
Read the full article here
