IntelligentHQ
IntelligentHQ
Skip to content

Enterprise AI & Agents

Software for building, customizing, deploying and governing AI assistants and agents in a company: Nemotron open models, NIM microservices, the NeMo libraries, AI Blueprints reference code, the OpenShell agent runtime and the AI Enterprise support layer.

Technology profiles for this category are in research.

Overview

Moving from a chatbot demo to agents that read documents, call tools and act on systems raises new questions: which model, how to ground it in private data, how to measure quality and how to stop an agent doing something it should not. This category groups NVIDIA software for those steps.

Nemotron is NVIDIA's open model family, with reasoning, retrieval, speech and safety models. NIM packages a model, an inference engine and standard APIs in a container. NeMo is a suite of libraries for data preparation, customization, evaluation, guardrails and agent tooling. AI Blueprints are reference applications on GitHub. OpenShell is an open runtime that sandboxes an agent and checks its file, network and model access against policy; its documentation lists version 0.1.2. AI Enterprise adds support and the production license that self-hosted NIM requires.

It suits platform teams whose data must stay in-house.

If a hosted model API meets your privacy, cost and latency needs, you may not need to self-host any of this. A small team testing an idea can start with a managed service and move later.123456

Problems it addresses

  • Grounding answers in company data1

    General models do not know internal documents. Nemotron Retriever models handle extraction, embeddings and reranking for retrieval-augmented generation.

  • Self-hosting models consistently2

    Every model needs its own serving setup. NIM bundles the model, an inference engine and APIs in a container that runs on NVIDIA GPUs in cloud, data center, workstation or edge.

  • Measuring and limiting behavior3

    Agents need tests and rules. NeMo Evaluator benchmarks models and agents, and NeMo Guardrails adds safety and compliance controls.

  • Containing what an agent can reach5

    Agents that run code can touch files and networks. OpenShell permits nothing by default and checks each request against policy outside the agent process.

  • Licensing for production2

    Self-hosted NIM is free for research and development through the Developer Program, but production use needs an AI Enterprise license.

A typical workflow

  1. Start from a reference4

    Begin with an AI Blueprint, such as the deep researcher agent or the Nemotron voice agent, instead of a blank project.

  2. Choose and serve the model1

    Pick a Nemotron tier (Nano, Super or Ultra) or another model and serve it as a NIM.

  3. Prepare data and customize3

    Clean data with NeMo Curator, generate synthetic data with NeMo Data Designer and fine-tune with NeMo Customizer if prompting is not enough.

  4. Evaluate and add guardrails3

    Benchmark with NeMo Evaluator and set rules with NeMo Guardrails before users see the agent.

  5. Run under policy5

    Execute the agent inside OpenShell so file, network and inference access follow written policy.

  6. Operate with support7

    License production deployments of NIM through NVIDIA AI Enterprise.

NVIDIA Solution Architect

Describe your project and get an explainable architecture.

Open the lab

Next steps

  1. Pick one narrow task with a measurable outcome, such as answering policy questions from a fixed document set.

  2. Try a Nemotron model through the hosted NIM APIs before deciding to self-host.

  3. Write the access policy for your agent (files, hosts, tools) before writing its prompts.

  4. Run the NVIDIA Solution Architect for a stack outline matched to your industry and data constraints.

Sources

  1. NVIDIA Nemotron foundation models (opens in a new tab)NVIDIA · Vendor-reported
  2. NVIDIA NIM Microservices product page (opens in a new tab)NVIDIA · Vendor-reported
  3. NVIDIA NeMo product page (opens in a new tab)NVIDIA · Vendor-reported
  4. NVIDIA AI Blueprints GitHub organization (opens in a new tab)NVIDIA · Vendor-reported
  5. NVIDIA OpenShell (opens in a new tab)NVIDIA · Vendor-reported
  6. NVIDIA OpenShell documentation (opens in a new tab)NVIDIA · Vendor-reported
  7. NVIDIA AI Enterprise product page (opens in a new tab)NVIDIA · Vendor-reported

Fill out the form below to request your copy.

Name(Required)