Luminary Adopts NVIDIA GeoTransolver AI Model Architecture, Unlocking Up to 10x Greater Accuracy in Physics AI

Luminary's SHIFT models retrained on new NVIDIA architecture deliver near-simulation-level precision at AI speed

Mar. 16, 2026 at 9:30pm

Got story updates? Submit your updates here. ›

Luminary, a Physics AI platform, announced it is the first to adopt NVIDIA's newest GeoTransolver model architecture for its SHIFT models, unlocking up to 10x greater prediction accuracy for engineering teams modeling real-world product performance. By retraining its entire SHIFT model portfolio on the new architecture, Luminary gives customers faster, more reliable design insights without the cost and delay of traditional computer-aided engineering simulations.

Why it matters

This collaboration between Luminary and NVIDIA aims to redefine how physical products are designed, as AI-driven simulation is transforming manufacturing industries. The adoption of the GeoTransolver architecture across Luminary's SHIFT models eliminates the traditional tradeoff between speed and accuracy, allowing engineering teams to access near-simulation-level precision at AI speed.

The details

Luminary's SHIFT models, built on the NVIDIA PhysicsNeMo framework, provide engineering teams with ready-to-use Physics AI models across industries like aerospace, automotive, and marine. By retraining these models on the new GeoTransolver architecture, Luminary is able to deliver significantly improved accuracy while maintaining the speed advantages of Physics AI. In internal benchmarks, the retrained models demonstrated near-simulation-level precision, allowing designs that once required weeks of computational simulation to now be evaluated in seconds with dramatically improved reliability.

  • Luminary announced the adoption of the NVIDIA GeoTransolver architecture on March 16, 2026.

The players

Luminary

A Physics AI platform that provides engineering teams with ready-to-use models built on the NVIDIA PhysicsNeMo framework.

NVIDIA

A technology company that defines AI infrastructure, including the NVIDIA PhysicsNeMo framework and the new GeoTransolver architecture.

Sceye

An aerospace company building lighter-than-air stratospheric platforms that is using Luminary's Physics AI to rapidly evaluate platform performance and refine designs.

Got photos? Submit your photos here. ›

What they’re saying

“Engineering teams have historically had to choose between speed and accuracy. By working with NVIDIA and adopting the GeoTransolver architecture across our SHIFT models, we are eliminating that compromise. Our customers now get near-simulation-level precision at AI speed, built directly into the workflows they rely on.”

— Suds Menon, Chief Product Officer at Luminary

“AI-driven simulation is redefining how manufacturing industries design and deliver increasingly complex products. By uniting NVIDIA PhysicsNeMo with Luminary's GPU-native solver, Luminary is opening new frontiers of intelligent engineering.”

— Tim Costa, VP and general manager of computational engineering at NVIDIA

“We are at the forefront of innovation in the stratosphere and we are proud to work with Luminary and leverage GeoTransolver to accelerate design exploration and optimization.”

— Sidharth Arunkumar, Design engineering manager, Sceye

What’s next

Luminary enables AI researchers, simulation engineers, or design engineers to immediately leverage the newly retrained SHIFT models and GeoTransolver architecture to start building enterprise-ready Physics AI solutions.

The takeaway

The collaboration between Luminary and NVIDIA to adopt the GeoTransolver architecture represents a significant advancement in Physics AI, eliminating the traditional tradeoff between speed and accuracy and redefining how physical products are designed across industries.