Dell Technologies Inc. today expanded the Dell AI Data Platform with a semantic layer and a knowledge graph intended to give artificial intelligence agents trusted context from a company’s own data.
Dell is also speeding up data processing on Nvidia Corp. graphics chips.
The new features are aimed at enterprise data that was never organized with an AI reader in mind. Much of it predates agents entirely. Dell said agents end up burning tokens on every query to piece together answers that should already exist.
Dell’s first fix is the Unified Semantic Layer, which gives every application the same business definitions and rules to work from, including any ontologies a company already maintains. An agent working across several systems no longer has to guess whether a “client” and an “account” are the same customer.
Alongside it sits the Enterprise Knowledge Graph, which tracks how a company’s data connects. When an agent asks a question, the platform draws on the graph to pull in every related table and vector index the agent is allowed to see, wherever the data lives. In Dell’s example, a manufacturer chasing a fault on a production line can trace one odd sensor reading all the way to the orders now at risk.
Knowledge Agents are built on top of the graph. Each one covers a single topic and works only from its assigned slice of company data. Customers decide what data an agent can see and how much it is allowed to spend. Nvidia’s Nemotron Retriever models handle reasoning and visual understanding for the agents. Because all three components hold some of a company’s most sensitive information, they run inside the customer’s own data center.
Arthur Lewis, president of Dell’s Infrastructure Solutions Group, said many enterprises have spent years making their data accessible without making it usable. An agent that can find a customer record but has no idea what it means or whether it can be trusted “isn’t intelligent,” he said. “It’s just fast.”
On the processing side, the Dell Data Processing Engine will run on graphics processing units using Nvidia’s cuDF library. Dell’s own September tests on a PowerEdge R770 server with Nvidia RTX PRO 4500 Blackwell Server Edition GPUs found Apache Spark jobs ran 3.9 times faster on average than on central processing units alone. The best result was a 20.4-times speedup on a batch data mining job, and the company noted that the tests used default settings with no tuning. Apache Arrow moves data between Dell storage and the engine so jobs can query it in place.
Dell PowerScale storage will support up to 500 tenants in a single cluster. Each tenant will get more granular role-based access control, and PowerScale will be able to encrypt and authenticate file traffic over the Network File System protocol with mutual Transport Layer Security. Dell is pitching the changes at AI service providers and enterprises running shared platforms.
A new Dell Storage Performance Tool benchmarks S3-compatible object storage across training, inference and checkpointing workloads to help customers size AI infrastructure. Dell is also expanding its implementation services for the platform to move customers from deployment into production.
The Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents are due in the first half of 2027. PowerScale’s multitenancy and security updates arrive in November. The accelerated Data Processing Engine follows in December, and more Apache Arrow acceleration is expected in the first half of next year. Customers can get the Storage Performance Tool and the new services today.