Blueprint Collection

The Blueprint Collection is a resource available on NVIDIA Build that provides developers with pre-built workflows and code samples designed to accelerate the construction of AI applications from the ground up. [NVIDIA Build Overview]

Blueprints lower the barrier to building production-ready AI applications by offering structured starting points — combining reference architectures, working code, and guided workflows — so developers can focus on customisation rather than foundational scaffolding. [NVIDIA Build Overview]

The Blueprint Collection sits alongside several other core capabilities on NVIDIA Build:

[NVIDIA Build Overview]

Developers are directed to Explore the Blueprint Collection from the NVIDIA Build homepage. A recommended entry point is setting up NemoClaw — NVIDIA's secure personal AI agent runtime — using the step-by-step playbooks also available through DGX Station. [NVIDIA Build Overview]

NVIDIA Build organises its offerings across four skill domains, all of which Blueprints may draw upon:

• AI and Machine Learning — 143 skills
• Physical AI — 37 skills
• Accelerated Computing — 25 skills
• Developer Tools — 10 skills

[NVIDIA Build Overview]

These domains are also reflected in the Inference Endpoints catalogue, where models such as those for Agentic AI, Physical AI, and MoE architectures are available for free inference. [NVIDIA Build Overview]

Alpamayo R1, NVIDIA's open reasoning vision language action model for autonomous vehicles, is paired with AlpaSim — a fully open simulation blueprint for high-fidelity autonomous vehicle testing. Partners including JLR, Lucid, Uber, and Berkeley DeepDrive are using AlpaSim to develop toward Level 4 autonomy. [NVIDIA AI Models 2026 Guide]

NVIDIA's Physical AI Stack introduces several blueprint-level resources tied to the Cosmos 2.5 world foundation model platform. These allow teams to generate synthetic training data across diverse conditions — covering robotics, autonomous driving, and industrial AI — before deploying to physical hardware. This dramatically reduces development cost by enabling simulation-first workflows. Adopters include Johnson & Johnson MedTech and Toyota Research Institute. [NVIDIA AI Models 2026 Guide]

Blueprints on NVIDIA Build may leverage NVIDIA NIM microservices for managed deployment of Nemotron 3 models, removing the need to manage GPU infrastructure directly. This integration is used in production by enterprises such as CrowdStrike (security AI), Bosch (in-vehicle voice via Nemotron Speech), and ServiceNow (Apriel model family trained on Nemotron datasets). [NVIDIA AI Models 2026 Guide]