Zaha Hadid Architects needed to shorten weeklong feedback cycles for highly detailed design scenes, protect a proprietary 40-year design archive, and scale AI-assisted visualization without sending sensitive client or intellectual-property data to external services.
Reported impact
ZHA reduced client feedback loops from one week to under 24 hours, fine-tuned custom LoRAs and Stable Diffusion models on proprietary design data, and created an extensible Omniverse-based platform that reuses decades of portfolio content while keeping sensitive design data on premises.
Zaha Hadid Architects built a local AI design pipeline around NVIDIA RTX 6000 Ada Generation GPUs, NVIDIA Omniverse, OpenUSD, and NVIDIA NIM large language models. High-end workstations with up to four GPUs handle rendering, simulation, model fine-tuning, and inference onsite so the firm can keep its 3D scenes, imagery, and client work local. ZHA combines Stable Diffusion, DALL-E 2, Rhino.ai, and custom LoRA workflows to turn parametric concepts into client-ready visuals faster, while NIM-backed tagging and search services organize its proprietary image archive for style transfer and geometry analysis. Through its CODE group, the firm also built more than ten Omniverse extensions and the ZSPACE Create interface so geometry processing, robotic simulation, environmental analysis, and rendering can share the same OpenUSD-based workflow instead of forcing teams to manually reformat scenes between tools.
Design geometry and proprietary image assets stay on local RTX workstations, flow through OpenUSD into Omniverse, are tagged and searched by NIM-backed services, and are then combined with generative models such as Stable Diffusion and DALL-E 2 to produce client visuals and design-analysis outputs.
They have developed over ten custom Omniverse extensions integrated into its Spatial Technology Stack.
Workstations equipped with up to four NVIDIA RTX 6000 Ada Generation GPUs deliver a high-performance local compute environment.
ZHA manages an extensive, proprietary image database using NVIDIA NIM large language models to tag, search, and retrain models.
ZHA reduced client feedback loops from one week to under 24 hours using AI-assisted pipelines and Omniverse.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Jul 28, 2026 Arup Arup needed to evaluate far more structural design options early in projects so engineers could optimize performance, material use, and embodied carbon, but traditional FEM simulations were too slow and queue-bound to… Up to 50% faster GNN model execution and training Graph neural network (GNN) surrogate models · NVIDIA DGX Spark NVIDIA Sample Feb 5, 2026 EDAG Production Solutions EDAG needed to help manufacturers launch and optimize factories faster, but fragmented planning tools, disconnected supplier data, and manual status tracking slowed collaboration, increased errors, and delayed readiness… More than 100 workers trained virtually before installation EDAG digital twin platform · NVIDIA Omniverse · OpenUSD NVIDIA Sample Feb 2, 2026 Worley Worley needed to unlock more than 50 years of engineering knowledge across over 100,000 documents, but fragmented tools, isolated repositories, and slow document retrieval caused engineers to waste time and made it hard… 100,000+ documents Microsoft Azure · Milvus · NVIDIA AI Enterprise 