MITRE has deployed several capabilities of the VAST Data Platform, with a new NVIDIA DGX SuperPOD to power the MITRE Federal AI Sandbox. This solution will provide researchers and developers across multiple federal government agencies with access to NVIDIA accelerated computing infrastructure and software for training large language models (LLMs) and experimenting with other generative AI tools to develop AI-enabled applications.
VAST Data Collaborates with MITRE and NVIDIA to Advance the US Federal Government AI Sandbox
Industry Government
Use Case AI
Overview
Background
The VAST Data Platform is built on VAST’s revolutionary Disaggregated Shared-Everything (DASE) architecture, and is the industry’s first enterprise-grade Network Attached Storage (NAS) solution certified for NVIDIA DGX SuperPOD, offering high performance data access and providing users parallel performance and exabyte scalability. MITRE researchers can also use the NVIDIA AI Enterprise software platform included with NVDIA DGX systems. Additionally, VAST’s implementation of the Data Pillar within the Zero Trust Architecture (ZTA) framework provides advanced security features and compliance with regulatory standards to create a secure, compliant environment for federal organizations to manage sensitive or unclassified data.
VAST Federal has a number of public sector organizations as customers already taking advantage of the VAST Data Platform – including the Department of Energy, NASA, National Oceanic and Atmospheric Administration (NOAA), National Institute of Health (NIH), and Homeland Security.
Outcome
With the VAST Data Platform, the MITRE Federal AI Sandbox provides federal agencies with:
Highly Performant, AI-Ready Data Infrastructure – VAST provides the performance, scale and a suite of data management capabilities to pursue AI-enabled applications for federal use cases like cybersecurity and critical infrastructure protection, high-resolution weather modeling and hazard planning, and modernizing and simplifying government systems.
Secure Multi-Tenancy – VAST’s multi-tenancy abilities allow MITRE to create and maintain multiple secure enclaves on a scalable data fabric. This helps federal agencies access and manage data using shared AI infrastructure resources without impacting performance or security. VAST’s Attribute-Based and Roles-Based Access Control (ABAC/RBAC) capabilities provide another layer of data security.
Improved Data Efficiency – As the MITRE Federal AI Sandbox continues to scale, VAST’s unique Similarity-based data reduction will help to further reduce the cost of data infrastructure and simplify data management as MITRE’s petabytes of data grow over time.
Support for Any Workload – AI requires random access across massive data sets, and VAST’s all-flash infrastructure and metadata capabilities allow MITRE to provide support for all workloads and AI pipelines, making the sandbox environment universally useful for diverse federal sponsor use cases and their AI deployments.
MITRE is working to reduce the barriers to AI adoption for federal agencies – and this new Federal AI Sandbox provides a modern accelerated computing environment capable of powering the many models, inferences and AI applications that federal agencies are working with today. We have incredibly talented engineers and researchers that are seeking new ways to manage and use data, networking, and supercomputing infrastructure - and this new cluster will open doors for the next-generation of AI-powered applications and research.
Charles Clancy
Chief Technology Officer & Senior Vice President, General Manager, MITRE Labs