Luminary Cloud aims to disrupt the CAE space

GPUs make near-real-time, accurate simulations the backbone to Luminary’s aims.

GPU- and cloud-based simulations have been on’s radar for quite some time. But a new CAE company, named Luminary Cloud, has come out of stealth aiming to finally disrupt the CAE space with this promising technology. The company calls its flavor of this tools real-time engineering and offers it via a CAE software-as-a-service (SaaS) platform. The package also comes rolled up with AI technology, in the form of an engineering design copilot, and cloud-based workflows that enable collaboration all around the world.

Simulation of a jet using Luminary Cloud. (Image: Luminary.)

Simulation of a jet using Luminary Cloud. (Image: Luminary.)

But what is most interesting is that this stealthy company has come into the light of day backed by interesting names. First there is its $115 million funding from Sutter Hill Ventures. Then there is a glowing review in press material from chief scientist and SVP of research at NVIDIA, Bill Dally. He said, “Success in virtual testing requires advanced accuracy, speed and versatility, and Luminary Cloud’s real-time engineering brings the industry closer than ever to that ideal … We’re very proud to make it possible for innovative new companies like Luminary to incorporate AI and accelerate computing to change the way more great ideas happen.”

Then there are Luminary’s customers, which include Joby Aviation, Mueller Co. LLC, Trek Bikes, Cobra Golf (a subsidiary of Puma), Piper Aircraft and Sceye. In other words, the technology has been tested to develop products for aerospace and defense, automotive, sporting goods, industrial equipment and more. Gregor Mikić, chief aerodynamicist at Joby Aviation, said: “With Luminary, we can quickly distinguish between ideas we should pursue, and those we should abandon. We are able to take complete aircraft configurations and run them in complex simulations in a matter of minutes, saving a significant amount of time. Luminary provides us with an entire set of tools that take us from geometry to result.” 

This news and praise, of course, spawned more questions than answers. So, reached out to Luminary to shed some light on its debut.

The accuracy of Luminary’s GPU-based simulation technology asked, how does the accuracy of Luminary’s technology compare to traditional CAE tools? For instance, Ansys Discovery focuses on near-real-time GPU-based simulations. But that product notes that these models are for product assessments that take place early in the development cycle. In other words, they lack the accuracy needed for a final product verification and certification. Is this true for Luminary? Or is Luminary boasting the speed of Ansys Discovery with the accuracy of Ansys Fluent, or similar?

Luminary says that it offers “both high performance and high accuracy, resulting from a brand-new, proprietary implementation of finite volume method solvers to perform well on pools of NVIDIA GPUs in the cloud. This is one of our significant product differentiators. We hired a multidisciplinary team of world-class computational physicists and computer scientists to implement arbitrary unstructured polyhedral physics solvers that have comparable-or-better accuracy to top-tier commercial solvers but can solve medium to high fidelity cases in minutes, rather than hours or days.”

“With the addition of Lumi Mesh Adaptation, a feature in our LUMI AI suite, we can produce superior and quantified accuracy to popular commercial solvers at high GPU-native speeds,” Luminary continues. “For example, Luminary can run steady RANS simulations with 150 million finite volume cells in minutes (7 minutes to be precise, for engineering accuracy, vs hours with existing tools). [It] also enables running transient DDES and WMLES simulations in under 30 minutes, thus enhancing the accuracy of the simulations even further when necessary.”

How smart is the AI-copilot Luminary has built?

Next, dug into the capabilities of Luminary’s AI copilot. AI-based simulations are becoming a trend in the industry. But you need to train these AI models on previous simulations so they can quickly and accurately predict performance. The question is, can Luminary’s AI tool do this? Or is it just used for pre- and post-processing? Finally, if it can run AI-based simulations, how much training does it need and is it pretrained in any way?

According to Luminary, “it has built a modern CAE SaaS product which addresses the entire CFD analysis workflow from pre-processing (CAD import and automated meshing) through simulation (fluid, thermal and porous media) to post-processing (3D scientific visualization, analysis and design exploration). Since our GPU- and cloud-native solvers already provide very high-accuracy physics simulation in seconds or minutes, the simulations require no training and are at least as accurate as popular commercial finite volume solvers.”

As for its AI features, Luminary adds, “We have focused our Lumi AI suite to be an engineering copilot for the tedious manual steps of the overall workflow. Currently we have targeted AI-assisted automated meshing and mesh adaptation as well as AI-assisted convenience features for things like runtime and cost estimation. There are a lot of manual steps in the conventional engineering analysis workflow which could be greatly improved by an AI copilot and Luminary is working on new capabilities across pre-processing, simulation, and post-processing.”

Will simulation democratization finally come to pass?

Simulation is often seen as an ivory tower where only specialists use the tools regularly and properly. So, wanted to know if this AI copilot aims to solve this issue. If so, how would it compare to a simulation expert making a simulation app for internal/external use as seen with the COMSOL App Builder?

Luminary answers that “it is early days for us in developing Lumi AI for democratization, but that is clearly the direction we are headed. In addition to Lumi AI, Luminary has invested a lot in ease of use, user experience and automation. For example, we have developed a ‘Library’ system that allows settings for meshing, boundary conditions, solver parameters, convergence criteria [and more]. [This library is] to be created by experts and used by others in their organization.”

Luminary adds that its SaaS platform “can automate ensembles of simulations and data management for designs of experiments and provide access to tabular data for surrogate modeling, all with clicks of the mouse or via Python scripts. We see the long-term opportunity in our real-time engineering vision as making CAE accessible to both advanced and ‘citizen’ analysts and enabling collaboration between the two.”

To that point, Luminary notes, “we apply this same philosophy to our commercial model as well: we embraced the consumption-based pricing model pioneered by Snowflake Computing in order to unlock the collaboration opportunity between CAE analysts and the tens or hundreds of design engineers that they collaborate with in their project teams. By having no per-user seat licenses, just pay-as-you-go usage-based computing, we enable accessibility by large teams of users: a CAE engineer can share their latest results with all of the design engineers on their team just by hitting ‘share.’ We will continue to enhance both Lumi AI and our workflow automation toward the goal of democratization and automated design exploration.”

So, will Luminary disrupt the CAE space? Since it’s only focused in CFD at this time, that’s unlikely. But as has said in the past, when disruption comes to engineering simulation, GPUs may take victory.

Written by

Shawn Wasserman

For over 10 years, Shawn Wasserman has informed, inspired and engaged the engineering community through online content. As a senior writer at WTWH media, he produces branded content to help engineers streamline their operations via new tools, technologies and software. While a senior editor at, Shawn wrote stories about CAE, simulation, PLM, CAD, IoT, AI and more. During his time as the blog manager at Ansys, Shawn produced content featuring stories, tips, tricks and interesting use cases for CAE technologies. Shawn holds a master’s degree in Bioengineering from the University of Guelph and an undergraduate degree in Chemical Engineering from the University of Waterloo.