NVIDIA AI Blueprints

Summary

NVIDIA AI Blueprints are NVIDIA-published reference workflows, code paths, and deployment patterns for building AI applications on NIM, NeMo, Nemotron, and the broader NVIDIA AI software stack. The build.nvidia.com blueprint catalog should be used as a discovery and current-state source, not as a reason to create a separate wiki page for every listed build or partner example.

Detail

Purpose

Blueprints give developers starting points for production-oriented AI applications. They commonly combine NVIDIA model endpoints or downloadable NIMs with orchestration code, deployment instructions, optional Helm or Docker paths, and guidance for connecting enterprise data or domain workflows.

Current scope

Agent workflows such as NVIDIA-AI-Q-Blueprint for enterprise research and deep research.
Retrieval workflows such as NVIDIA-RAG-Blueprint for enterprise RAG, multimodal retrieval, evaluation, and guardrails.
Continuous model and agent optimization workflows such as NVIDIA-Data-Flywheel-Blueprint, with NeMo-Data-Designer, NeMo-Customizer, NeMo-Evaluator, and NeMo-Auditor providing data generation, adaptation, measurement, and safety-audit services.
Video intelligence workflows such as NVIDIA-Video-Search-and-Summarization-Blueprint for vision agents, search, summarization, and alert verification.
Digital-human workflows such as NVIDIA-Tokkio-Digital-Human-Blueprint for avatar-based customer-service and enterprise assistant experiences.
Domain application examples across voice agents, healthcare agents, retail assistants, telecom, warehouse operations, financial data, and physical AI data search.
Physical AI and multimodal workflows can use NIM-for-Cosmos-Embed1, NIM-for-Vision-Language-Models, and NIM-for-Visual-Generative-AI without turning every build.nvidia listing into its own wiki page.
Deployment paths that can include build.nvidia.com endpoints, NVIDIA-Brev cloud environments, NGC artifacts, Docker Compose, or Kubernetes/Helm depending on the blueprint.
Red-Hat-AI-Factory-with-NVIDIA is an OpenShift AI deployment foundation for validated NVIDIA Blueprints and agent workflows in a Red Hat environment.

NVIDIA context

Blueprints sit between NVIDIA developer samples and enterprise production platforms. They show how NVIDIA-NIM, NeMo-Platform, NVIDIA-Agent-Intelligence-Toolkit, NeMo-Retriever, Nemotron, current model choices such as Nemotron-3-Nano, Nemotron-3-Super, and Nemotron-3-Nano-Omni, and NVIDIA-AI-Enterprise fit together in real application patterns.

Wiki curation guidance

Do not create one wiki page per blueprint listing. Create a page only when a blueprint is a durable NVIDIA-authored topic with its own stable docs, architecture, or ecosystem significance. Smaller or duplicate examples should be folded into the canonical topic page, such as Nemotron, NVIDIA-Riva, NVIDIA-Cosmos, NeMo-Retriever, or NVIDIA-AI-Q-Blueprint.

Connections

NVIDIA-NIM - many blueprints rely on hosted or downloadable NIM model endpoints.
NIM-for-Large-Language-Models - LLM NIMs provide the self-hosted model serving path for many generative AI blueprints.
NVIDIA-NIM-Operator - Kubernetes deployment path for blueprint stacks that include NIM/NeMo microservices.
Red-Hat-AI-Factory-with-NVIDIA - OpenShift AI deployment guide that uses validated blueprints as a path from agent prototypes to production.
NeMo-Retriever-Embedding-NIM and NeMo-Retriever-Reranking-NIM - retrieval NIMs commonly appear in RAG and agent blueprints.
NIM-for-Cosmos-Embed1 - durable NIM for physical AI video search and dataset curation workflows referenced by blueprint-style applications.
NIM-for-Vision-Language-Models and NIM-for-Visual-Generative-AI - multimodal understanding and visual generation NIM families that can compose with blueprint workflows.
NVIDIA-NemoGuard-NIMs and NIM-for-Multimodal-Safety - safety NIM families that can moderate text and multimodal blueprint workflows.
NVIDIA-AI-Enterprise - enterprise blueprints map to NVIDIA-supported production AI software.
NVIDIA-RAG-Blueprint - durable RAG Blueprint for enterprise retrieval, multimodal RAG, and knowledge-grounded generation.
NVIDIA-AI-Q-Blueprint - durable AI Blueprint for enterprise research agents.
NVIDIA-Data-Flywheel-Blueprint - durable AI Blueprint for continuous agent and model optimization.
NVIDIA-Video-Search-and-Summarization-Blueprint - durable VSS Blueprint for vision agents and video analytics.
NVIDIA-Tokkio-Digital-Human-Blueprint - durable Digital Human Blueprint for interactive avatar experiences.
NeMo-Platform - blueprints often use NeMo microservices for customization, evaluation, retrieval, and inference.
NeMo-Data-Designer, NeMo-Customizer, NeMo-Evaluator, and NeMo-Auditor - NeMo Platform services that turn blueprint workflows into data, adaptation, evaluation, and safety-audit loops.
Nemotron, Nemotron-3-Nano, Nemotron-3-Super, and Nemotron-3-Nano-Omni - current blueprints use Nemotron models for reasoning, speech, multimodal perception, and agentic workflows.
NVIDIA-Brev - some blueprints expose launchable cloud development paths through Brev.

Source Excerpts

build.nvidia.com lists NVIDIA and partner blueprints for AI applications, with tags for launchability, enterprise use, NIM, NeMo, Nemotron, and domain categories.
The AI-Q and Data Flywheel blueprint cards show how durable blueprints can become canonical wiki pages when they have stable NVIDIA-authored architecture and deployment surfaces.

Resources

Build NVIDIA Blueprints

AIPS BOOM

Explorer

NVIDIA-AI-Blueprints