Everpure has unveiled new data-management capabilities to help enterprises move AI workloads from pilot projects into production, focusing on data governance, inference performance and cost control.
The updates address three barriers to scaling enterprise AI: fragmented data context, unpredictable inference costs and complex deployments.
“Enterprise AI is hitting a wall not because the models are lacking, but because data is not ready for real-time, autonomous agents,” said Prakash Darji (top), General Manager of Data and Digital Experience at Everpure. “We are eliminating that friction. By making enterprise data continuously governed, automated, and instantly accessible, we’re giving organisations the foundation to move AI out of the lab and into production with the necessary confidence.”
Everpure Data Intelligence is designed to discover, classify and contextualise information across the Everpure Platform, public clouds, software-as-a-service applications, and third-party storage.
New capabilities include native integration with the open Model Context Protocol that lets AI agents and security tools query live data catalogues using natural language while identifying relevant information and its sensitivity classification.
The company also announced turn-key deployment through its existing Pure1 console, which should reduce the need for separate management servers and extensive professional-services engagements.
A privacy-focused file-intelligence capability will show which users can access file shares and how stale the data is without reading file contents. This would help enterprises address exposure risks and reclaim storage capacity before making data available to AI agents.
Faster inference, lower costs
Everpure is also adding PureKVA, or Key-Value Accelerator, to its FlashBlade platform. The technology pre-stages context directly into GPU memory and is designed to deliver up to 20 times faster time to first token, according to the company.
Supporting multi-tenant deployments without relocating datasets, PureKVA reduces GPU idle time, increases token throughput and lowers response latency for real-time applications.
Also introduced is Always-On DeepReduce, which continuously scans storage blocks across FlashBlade systems to identify sub-block similarities that conventional deduplication may miss, including in pre-compressed data. It can expand usable capacity automatically without affecting write performance or requiring manual scheduling, potentially reducing hardware and cross-cloud costs.
A new reference architecture based on open-weight models helps enterprises retain greater control over their data while reducing reliance on external application programming interfaces. It can lower overall API token usage and make AI spending more predictable.
