Cloudian has updated its HyperScale AIDP product with native ingestion from file- and object-based storage sources, Nvidia blueprint support and data access permission controls.
The HyperScale AI Data Platform (AIDP) is a turnkey, on-premises appliance that’s an alternative to public cloud S3 data sources. It’s an S3-compatible, RDMA-based, data store for AI models and agents running on Nvidia Blackwell GPU hardware and software and BlueField DPUs, and complying with Nvidia’s AI Data Platform reference architecture. It lets customers run production AI on their own IT infrastructure, keeping sensitive data under their direct control. Cloudian says it saves up to 60 percent on cost by avoiding the recurring token, egress, and inference fees that make public cloud AI expensive at scale.
Cloudian CTO Neil Stobart said: “Enterprise AI succeeds or fails on data access. Customers tell us their projects stall not because the models lack capability, but because the data those models need is scattered across file and object silos, or because opening up that data would violate internal access policies. HyperScale AIDP v1.1 addresses both barriers directly. It reaches across the data estates customers already have, and it carries permissions all the way through to inference — so the right people get the right answers, and nothing more.”
HyperScale AIDP v1.1 enables AI apps running in the GPUs to access data from from NFS file shares and S3-protocol object stores simultaneously without consolidating or migrating data first. HyperScale AIDP indexes and vectorizes content without making a second copy, storing only the resulting embeddings.
The updated HyperScale AIDP enforces user- and group-level access controls throughout the ingest, indexing, retrieval, and response section software the AI data pipeline. Each AI query returns only the content the requesting user is authorized to see. Those controls extend to the vector database itself, blocking unauthorized access to the embeddings.Cloudian says this is an important extra safeguard because exposed vectors can be inverted to reconstruct the underlying source data. This is known as an embedding inversion attack.
The v1.1 update means that HyperScaleAIDP now supports Nividia’s AI Blueprint for Enterprise Document RAG and Metropolis Blueprint for video search and summarization (VSS).
Enterprise Document RAG transforms an organization’s existing volumes of contracts, policies, technical documentation, and reports into a conversational knowledge base that returns precise answers with source citations. Metropolis VSS Blueprint extends the same capability to video, generating responses from live and recorded visual sensors, training content, broadcast archives, and inspection videos.
VSS automatically indexes new footage as it arrives rather than doing it through scheduled batch processing. This makes VSS practical for security/compliance monitoring, quality control inspection, and other use cases where video accumulates in real time. To cite obvious examples, facial recognition and visual product quality checks can be run in real time.
Both these blueprints run natively with HyperScale AIDP’s S3-native storage and built-in Milvus vector database.
HyperScale AIDP is validated on Nvidia-Certified GPU server platforms from both Supermicro and Lenovo. It’s available now through Cloudian and its global partner network.
There is a demo of the VSS continuous ingest feature here and an additional VSS demo with a different use case here.