VAST unveils Foundation Stacks to speed AI to production

VAST Data has launched Foundation Stacks, an open-source library that packages NVIDIA AI Blueprints into production-oriented pipelines designed to run on the VAST AI Operating System.

The library targets organisations looking to move AI projects from prototype to production. Many enterprises have adopted reference patterns for AI applications, but still face significant integration work before those examples can run securely and reliably at scale.

Foundation Stacks build on NVIDIA AI Blueprints, which provide starting points for developers building AI applications and intelligent agents. NVIDIA positions the blueprints as a way to prototype and customise domain-specific workflows using NVIDIA AI Enterprise software.

VAST's approach focuses on templates that run natively on its platform, letting developers spend more time on application and business logic instead of assembling infrastructure components.

Production templates

Enterprises often need to connect multiple layers of infrastructure and services before a reference AI workflow can run in production. That work can include compute orchestration, data services, and operational tooling for reliability and security.

Foundation Stacks are repeatable implementations designed to reduce that integration burden by bundling multiple elements in one environment. VAST cites data access, database services, compute orchestration, eventing, and pipeline execution as the building blocks unified under the VAST AI Operating System.

They also support a common goal for large organisations: standardising AI deployment patterns so teams can apply the same approach across projects and business units. The stacks can be deployed repeatedly wherever the VAST AI OS runs.

VAST is positioning the library for both cloud and on-premises environments, including deployments on its CNode-X platforms, which it has presented as part of the NVIDIA AI Data Platform reference design.

First releases

The first Foundation Stacks are based on two NVIDIA AI Blueprints: Video Search and Summarization and NVIDIA AI-Q. VAST describes each as a production-ready implementation running on the VAST AI Operating System.

The Video Search and Summarization-based stack focuses on video ingestion and analysis across live and archived footage. It supports semantic indexing, summarisation, and interactive Q&A over video.

The AI-Q-based stack targets agent-style workflows operating across private enterprise data sources. VAST says it provides a foundation for building custom AI researchers, with features including persistent context, scalable reasoning pipelines, and what it calls trusted agent execution.

John Mao, Vice President, Global Technology Alliances at VAST Data, said the release bridges reference designs and production deployments.

"NVIDIA AI Blueprints have given the market an important starting point for building next-generation applications, but enterprises still need a production-ready way to deploy and operate those capabilities at scale," Mao said. "With VAST Foundation Stacks, VAST is taking the architectural patterns behind leading NVIDIA Blueprints and giving customers a faster path from experimentation to production for scalable AI pipelines, video intelligence, and agentic AI systems."

NVIDIA linked the announcement to a broader challenge in production AI programmes: keeping enterprise data continuously ready for models and agents. Adel El Hallak, Vice President, Product at NVIDIA, pointed to the need for integration across infrastructure layers.

"As enterprises transition to production AI at scale, preparing enterprise data for AI has become one of the biggest challenges," El Hallak said. "Turning data into AI-ready pipelines needs to be done continuously and requires full-stack acceleration across compute, networking and software. By extending NVIDIA AI Blueprints with the VAST AI Operating System, customers can prepare and serve data so intelligent agents are always working off the most recent and accurate information."

Availability and roadmap

Foundation Stacks will be distributed through a public GitHub repository. VAST also referenced interactive demos and planned sandbox environments for customers and partners.

VAST plans to publish additional stacks in the coming months, including examples tailored to specific industries.