AI is taking shape, moving from models that answer questions to models that can reason. Now models are powering the agentic era, and Dell Technologies Inc. has announced several extensions to its AI Data Platform designed to manage this new paradigm.
These include extending the Dell Data Orchestration Engine with three new capabilities: the Unified Semantic Layer, the Enterprise Knowledge Graph and Knowledge Agents. Dell also added a new capability in the Data Processing Engine designed to accelerate the enterprise data prep for AI. The engine will run on graphics processing units using Nvidia Corp.’s cuDF library.
The intent, as explained on theCUBE, SiliconANGLE Media’s livestreaming studio, by Arthur Lewis (pictured), president of the Infrastructure Solutions Group at Dell, is to bring the Data Processing Engine and Dell’s storage engines closer together, reducing unnecessary data movement with capabilities such as zero-copy data access, memory-to-memory transfer and distributed processing. Dell is creating a more efficient data path from raw enterprise information to AI-ready context, key elements of the AI Data Platform and the Dell AI Factory.
“In many cases, the model is not the next constraint, the data is,” Lewis said. “That is the problem that the AI Data Platform was built to solve. Within the Dell AI Factory, it is the data foundation that curates, connects and serves enterprise data to fuel the AI, and especially agentic AI.”
This feature is part of SiliconANGLE Media’s exploration of the architectural shifts powering continuous, production-grade AI. Be sure to check out theCUBE’s “Dell AI Data Platform Event: From Ambition to AI at Scale.” (* Disclosure below.)
Storage solutions for the agentic era
Dell’s announcements underscore the importance of storage in different stages of the AI lifecycle. In bringing the Data Processing Engine closer to Dell’s storage architecture, the company is leveraging the capabilities of PowerScale, ObjectScale and Lightning File System to target different performance, capacity and workload requirements.
PowerScale is the foundation for enterprise AI at scale, with support capabilities for up to 16,000 GPUs. ObjectScale provides customers with the object foundation for enterprise data, delivering up to 40 gigabytes per second per node. According to Dell’s AI Storage Product Management team, Lightning is the world’s fastest parallel file system, handling workloads that push into extreme throughput, large-scale training, checkpointing and the most demanding inference paths.
At AI cloud scale, storage becomes part of the underlying business model, with its performance affecting revenue per GPU, infrastructure utilization and how quickly customers can move into billable production.
Dell is positioning its AI Data Platform as the connective layer between enterprise data and production AI, bringing together storage, data processing, orchestration, search, governance and GPU acceleration. Today’s announcements essentially connect the data orchestration engine, storage semantic layer, knowledge graph at scale and cyber resilience into one AI data architecture.
“They’re focusing on helping customers get AI right,” said theCUBE Research’s Dave Vellante. “Everybody wants to get AI right. That means getting to the right information, making sure that information is harmonized and ready, and then delivering it efficiently at a low cost with low operational complexity. Those are the big things. Two years ago, we said, look, the enterprise needs solutions. And what Dell is doing is they’re bringing a solution.”
Extensions for semantic layers and knowledge graphs
A key part of Dell’s solution moving forward will involve newly announced extensions to the Dell Data Orchestration Engine with a Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents.
The Unified Semantic Layer offers an enterprise-wide knowledge graph native to the Data Orchestration Engine. Dell is seeking to address frustration on the part of customers who have worked with other semantic tools that reach only structured data sources, leaving unstructured information behind. Another source of friction is that other offerings build context once and rely on users to keep it updated as the underlying data evolves.
“We’re doing it differently,” said Vrashank Jain, lead product manager, AI Data Platform at Dell, during an interview on theCUBE. “The semantic layer gives structured and unstructured information consistent business meaning, so a term means the same thing everywhere it appears. It uses open models to fix the hardest part of building semantic layers: generating entities, definitions and finding meaning at-scale with human-in-the-loop for verification. No more standardized ontologies. Each organization is unique. Their semantic layer should reflect that.”
Dell’s Knowledge Graph connects entities and relationships so an application or agent can understand how data is related. It uses metadata from structured and unstructured sources, lineage, storage metadata and query history to continuously tune the graph. Once an agent asks a question, the platform augments the context surrounding the query with related entities such as tables, images, vector indexes and logs, according to Jain.
Dell’s Knowledge Agent is a new way of interacting with the AI Data Platform, powered by the Unified Semantic Layer and Knowledge Graph. After building the graph and semantic layers, users will be able to develop Knowledge Agents focused on a part of the Knowledge Graph that is curated with a specific expertise. Users can give the agents prompt-based guidance, data access permissions, quality guardrails, token cost thresholds, all while remaining model agnostic.
“A Knowledge Agent is like a trusted advisor on your team who is an expert on any given topic,” Jain told theCUBE. “It brings trust and accuracy to agentic retrieval.”
Enhancing the Dell AI Data Platform
The announcements from Dell reflect several key transformations currently taking place in the world of enterprise compute.
One involves an evolution of the AI stack itself, brought on by the rapidly expanding influence of agents. Enterprises now want to get to agent outcomes as fast as possible. This will involve connecting the underlying enterprise data and converting it into context that can then be served to agents.
“We should step back and look at how we are moving from these so-called AI factories or intelligence producing factories, which they were originally conceived as, to what we are now calling agentic data centers,” said Gaurav Chawla, vice president and Dell fellow, Infrastructure Solutions Group, at Dell, in conversation with theCUBE. “And when you look at an agentic data center, it’s not just the model, it’s not just the storage, but you need a combination of models, the storage and the data stack.”
Dell is also retooling how it ultimately serves enterprise customers, transforming its business to operate at the speed and agility of an AI-native company. This is a major shift from what we have seen so far, according to Chawla, because AI is becoming much more than just a copilot for its human operator.
“This is a shift now evolving from human as the operator [to] where you want the agent to be the operator and agents bring humans back in the loop when they need them to review or to approve something,” Chawla said. “If architected and deployed correctly, this means agents not only execute business processes, but they also learn and get better over time. So that’s what we are building at Dell AI Data Platform.”
The various elements of this transformation point toward an important shift in the role of the AI Factory. As theCUBE Research’s John Furrier has noted, the data center is becoming an industrial system in its own right, and the AI infrastructure race is no longer just about compute; it’s about data movement.
“My big takeaway is that the most important thing that’s happening here is two things: Dell has a new data architecture that’s unified and two, the industry itself is moving to this phase where … data centers are going to become industrial centers, multiple data centers working together,” Furrier said. “The story here is that the data path is the new architecture.”
Stay tuned for the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Dell AI Data Platform Event.
(* Disclosure: TheCUBE is a paid media partner for the Dell AI Data Platform Event. Neither Dell, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)