Generally available capabilities in the VAST AI Operating System make Enterprise AI at scale a reality
Enterprises increasingly seek to leverage AI and advanced analytics to drive innovation and operational efficiency. However, persistent challenges - such as inflexible legacy systems, fragmented toolsets, and critical concerns around data governance and security - continue to impede scalable AI adoption.
The GA of the latest VAST AI Operating System version directly addresses these challenges. As the only platform for end-to-end real-time operational AI, the VAST AI OS integrates real-time vector storage and search with event-driven processing for automated pipelines, all within the VAST DataEngine. VAST InsightEngine builds on these capabilities to deliver a unified pipeline builder and orchestration layer for transforming raw data into intelligence. All of it is wrapped in comprehensive, identity-based governance from raw data to vector.
This unified ecosystem also supports complex analytics at scale. The VAST DataBase now delivers enhanced performance through its new Sorted Tables feature, making near-logarithmic query speeds across massive datasets possible.
With today’s release, the VAST DataBase has evolved into an AI-native, lakehouse-grade platform featuring real-time vector search and enterprise-grade governance. And the VAST DataEngine now powers next-generation, event-driven AI pipelines and data operations - all running on a unified, all-flash DASE architecture.
Let’s examine these capabilities in detail.
The DASE Architectural Foundation
At the foundation of everything VAST is the Disaggregated Shared Everything (DASE) architecture. DASE fundamentally decouples compute resources (CNodes) from persistent storage (DNodes), ensuring that all CNodes have direct, high-speed NVMe-over-Fabrics access to the entire global dataset. This shared-everything model eliminates data movement bottlenecks, avoids the complexities of traditional data sharding, and enables massive parallelism for all workloads. This architectural design is critical for delivering the platform's real-time vector transaction performance, enabling serverless functions to execute directly on data, and allowing analytics engines to operate with unparalleled efficiency and scalability.
Accelerating AI: From Raw Data to Real-time Intelligence with End-to-End Governance
AI success depends on access to real-time, high-quality data and the ability to transform and govern it at every step. The VAST DataEngine provides a serverless compute environment for building flexible workflows and running lightweight, event-triggered functions.
Built on the VAST DataEngine, VAST InsightEngine is a powerful orchestration layer for building unified, automated, and governed AI pipelines from data ingestion to inference. Unlike traditional approaches that require manual handoffs between tools, InsightEngine provides a seamless, event-driven framework for turning raw data into intelligence in real time.
As data lands in the platform, InsightEngine pipelines automatically trigger lightweight, serverless Python functions - executed as stateless containers on VAST CNodes - to enrich, transform, and vectorize content. These functions can parse unstructured documents, apply metadata extraction, integrate with external services (like NVIDIA NIMs for model-based embedding), and update downstream systems, including the vector-native VAST DataBase.
What sets InsightEngine apart is its ability to combine real-time data processing, AI enrichment, vectorization, and inference within a governed pipeline. When inference is required, InsightEngine supports fully automated RAG pipelines, with integrated real-time Vector Storage and Retrieval and native SQL querying.
With unified AI governance, every stage of the workflow - from raw data to vector - is secured with identity-based access control. This ensures that only authorized users and models interact with permitted data, enabling organizations to build AI applications that are not only fast and intelligent, but also fully compliant and secure.
Next-Gen Analytics: Unifying Data for Faster Insights and AI Readiness
Traditional analytics stacks are slow, complex, and hard to scale. The VAST AI OS simplifies everything with high-performance distributed SQL directly on the platform with no data movement required.
Now GA, Sorted Tables for Log-Time Search feature unlocks near-logarithmic performance on massive datasets. In real-world scenarios, we’ve seen point queries improve by up to 100x and Trino queries on 10B+ rows speed up by 95x—reducing latency from 12 seconds to just 144ms.
A New Model for Operational AI
This release marks a fundamental shift in what’s possible for enterprise AI. With real-time vector search, serverless pipelines, high-speed analytics, and end-to-end governance on a single platform, VAST enables organizations to simplify infrastructure and unlock the full value of their data.
Key benefits include:
Real-Time Operational AI: Vector-native search and event-driven pipelines on live data
Accelerated Analytics: Faster queries, serverless functions, and automated enrichment
Unified Governance: Fine-grained identity-based controls from ingest to insight
Simplified Architecture: One unified, scalable platform for structured, unstructured, and vector data
Join us for our VAST Live session, "Product Update: Unlocking Real-Time Operational AI" on October 14th at 9am PT. We'll show you how our latest advancements are revolutionizing data analytics and helping you unlock new possibilities with AI.
You also can book a demo today or jump into the conversation on Cosmos.
