BMW Group needed to scale industrial AI across quality assurance, production, and logistics, but limited training data, manual image labeling, and a lack of end-to-end lifecycle tooling slowed model development and made it hard for plant teams to deploy AI autonomously.
Reported impact
BMW achieved an 8x boost in data-science productivity, 4–6x better performance than legacy systems, cut employee time to implement QA automation by more than two-thirds, and can generate hundreds of thousands of synthetic images on demand to support production AI at scale.
BMW Group built a complete deep learning operations pipeline on NVIDIA DGX systems so manufacturing teams could generate training data, train models, and deploy AI applications from one platform. Using NVIDIA Omniverse Enterprise, the company creates virtual factories and synthetic production scenes that feed the SORDI dataset, the open industrial reference set BMW uses for object detection, segmentation, and related vision tasks. DGX systems with Hopper architecture then train synthetic-data generators and downstream quality models, while TAO inside NVIDIA AI Enterprise supports AutoML and inference for no-code applications such as LabelTool Lite. This lets employees load a few examples, enrich them with synthetic images, and quickly build visual-inspection models for tasks like door-sill validation and stitch checking. The result is an industrial AI stack that reduces manual labeling, speeds deployment, and makes advanced quality-automation tooling usable well beyond the R&D team.
Synthetic factory scenes are generated in Omniverse, converted into large labeled datasets such as SORDI, trained on DGX systems, and then pushed into no-code and production QA workflows where plant employees deploy inspection models with minimal manual labeling.
DGX systems delivered an 8X boost in our data scientists’ productivity
SORDI ... comprising over 800,000 photorealistic images spanning 80 categories
The BMW Group is using NVIDIA Omniverse to create virtual factories and simulate complex scenes, along with NVIDIA DGX systems to create synthetic data based on those simulations.
The time it takes for our employees to implement AI automation in quality assurance has been slashed by over two-thirds.
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 Aug 11, 2025 Magna Magna needed to improve efficiency, safety and agility across its global manufacturing operations while also helping its autonomous delivery vehicle platform adapt to new cities and complex physical environments. City Delivery platform · NVIDIA Cosmos Reason · NVIDIA Omniverse simulation APIs and libraries NVIDIA Sample Archived · Dec 10, 2023 BMW Group BMW Group needed to plan and reconfigure large factories years before production starts, but disconnected CAD tools, manual coordination, and expensive change orders made it hard for planners to collaborate and validate… Projected 30% savings from optimized factory planning and highly efficient processes FactoryExplorer · NVIDIA Omniverse · Omniverse Kit SDK 
In-house Sample Oct 16, 2026 Mercedes-Benz Mercedes-Benz needed to digitalise global product development, connect development, procurement, production, sales, and after-sales data end to end, and help teams make better decisions earlier so new products and… efficiency gains of up to 50 percent expected in individual engineering processes AI agents · CONTACT Elements · PLM.AI 