Dell Technologies Dell Deskside Agentic AI Solutions for NVIDIA

Dell Technologies  ·  AI Factory with NVIDIA

Start building enterprise AI on Dell deskside systems

Dell-opinionated reference architecture solutions for NVIDIA-equipped Dell Pro Precision workstations — with S/M/L sizing, runnable serving recipes, and aiperf-validated performance.

Browse Solutions Sizing & Stack

Dell-opinionated reference solutions for enterprise AI on Dell deskside systems with NVIDIA technology.

Get Started

Start here — what Dell Deskside Agentic AI Solutions delivers and how to use it.

OverviewWhat this program delivers and how the solutions fit together. Solution CatalogAll Dell-opinionated solution templates at a glance.

Solution Templates

Opinionated, open-source reference architectures. Each ships S/M/L sizing, a runnable recipe, and validated benchmarks.

Agentic CodingPrivate on-deskside AI coding assistant behind an OpenAI-compatible API. Agentic Video Search & SummarizationSearch, ask questions over, and summarize local video libraries on Dell deskside GPUs. OpenClaw on NemoClawThe OpenClaw assistant on NVIDIA's NemoClaw runtime — tool-using agents with policy and sandboxing, local LLM. Hermes on NemoClawThe shipping tool-using agent on the NemoClaw runtime, fully local. Higher Education AI LabRepeatable design for university AI teaching labs on Dell hardware.

Deployment & Reference

The technical reference a builder needs to deploy a solution: recipes, sizing, and the inference stack.

Deployment RecipesRunnable serving recipes per platform and server. S/M/L SizingMap Dell system → model → inference config → performance. Inference Stack & ModelsvLLM-preferred stack and per-tier, per-workload model recommendations (Lightning on S; Super VL on M/L; tier-scaled NVFP4 alternatives). Stack CompositionHow the layers of the solution stack compose. PersonasIT Admin, functional SME, and AI engineer. GlossaryTerminology and acronyms. Architecture BuilderInteractively customize the reference architecture per use case; export SVG/PNG/HTML. Topology (TP vs DP)Tensor parallel vs data parallel, measured per platform — a per-workload choice, not a stylistic one. Scope & BacklogWhat this program covers today, and the design directions it supports next. OperationsPort map, co-tenancy rules, secrets, airgap loading, and the operator FAQ. Tokenomics MonitoringPrometheus + Grafana token-throughput stack that comes up with every recipe. ChangelogRelease history for the templates, recipes, and measured data.