Luminary Cloud Takes CAE to the Cloud

April 19, 2024

Sebastian Blanco

The cloud is many things to many people, but it’s really just a new way to talk about using other people’s computers instead of your own. When it comes to expensive GPUs that run intensive simulations, using someone else’s computers — theirs — is exactly what Luminary Cloud wants the automotive industry to do.

Luminary Cloud came out of stealth mode in early March, just in time to make a splash at NVIDIA’s GTC 2024 event. With the support of $115 million in funding from Sutter Hill Ventures, Luminary Cloud now wants to offer the “world’s first modern computer-aided engineering (CAE) software as a service (SaaS).”

Luminary Cloud provides CAE simulation solutions of various physics, with all the physics solvers being based on GPU computing. The SaaS portion of the equation uses NVIDIA’s GPUs. It can “share massive amounts of data, large simulations with terabytes of data, with just a share button, like a Google Doc,” Luminary Cloud, Inc. CTO and co-founder Juan Alonso told SAE Media. “I can distribute it across my entire team. That comes from the fact that all your data is stored, and it’s always available anytime, anywhere, in a secure way.” Alonso said customers can also write their own Python scripts to use with the service and create new models to infuse into their workflows. Customers do not need to install any software or maintain a cluster to use Luminary Cloud’s “real-time engineering” services.

The computer-aided engineering (CAE) part comes from Luminary Cloud’s physics software, which can ingest CAD files, generate meshes, run solutions and extract solutions for visualization.

“One of the things that we’ve done is, by leveraging the GPUs in the cloud, we give you elasticity,” Alonso said. “We can run one simulation, but 100 times faster than the competition, so instead of taking four to six hours to do a RAN simulation on a hundred million cells, we can actually do it in a couple of minutes. If you’re in a crunch to run, let’s say, 1,000 simulations in a day, you don’t have to have a cluster that has 30,000 GPUs to do that, which would be a waste for your company. Rather, you can utilize them for one day and then move on.”

24 hours in 60 minutes

Ian Lockley, Luminary Cloud’s principal solutions engineer, told SAE Media that the company’s physics simulations impressed a potential automotive customer who asked them to run a blind test analyzing a transient aeronautics simulation. The customer’s previous effort to run the simulation took them 24 hours.

“Now, in my job, you never want a blind test,” Lockley said. “You always want the test data before you run the simulation. But they said, ‘Run a simulation of this vehicle, and we’re not gonna tell the answer. ’So we ran that simulation in less than an hour, and we were within 3% of the results of that test data. So that was pretty cool.”

Alonso said Luminary Cloud offers another advantage over other cloud-based CAE companies. The company knew it wanted to deliver speed but soon realized that running simulations in less time requires more than just better processors.

“Speed’s not just speed of the solver,” he said. “It’s the end-to-end process, importing the CAD, actually cleaning up the CAD and making the meshes, possibly adapting the meshes, so you don’t have to have a high degree of expertise in order to get accurate calculations and then interpreting the data.

“The initial premise of the company was to bring the solution time from hours to minutes. And early on in the company, we realized we could do that. Then we saw that setting up the visualization takes a long time, interpreting the results takes a long time, bringing in the CAD, setting up the mash, and realizing the mesh is not refined enough in various places. So we’ve been automating all of the steps in that process so that, eventually, people with less expertise can actually run Luminary and have confident and very quick results.”

Beyond flow physics

Luminary Cloud will announce some of its automotive industry partners later this year, and Alonso said he expects case studies to be published as well.

“We are quite interested in expanding beyond flow physics or aerodynamics,” he said, adding that porous media and conjugate heat transfer simulation capabilities will be announced soon. “As you can see, we are trying to work on the complete solution for automotive,” he said.

Alonso said Luminary Cloud grew from one pre-paid customer at the start of 2023 to 33 at the end of the year, along with “a number” of pay-as-you-go customers. The company also grew from about 45 people to almost 100 in 2023.

Alonso said Luminary Cloud is taking a page from its sister company, Snowflake, when it comes to operating entirely as a pay-as-you-go, cloud-based data SaaS. Mike Speiser, Sutter Hill Ventures MD and the founding CEO of Snowflake and Pure Storage, serves on Luminary Cloud’s board. Alonso founded Stanford’s Aerospace Design Laboratory and was formerly the director of NASA’s Aeronautics research program. Luminary Cloud’s other co-founder, CEO Jason Lango, was an Entrepreneur in Residence at Sutter Hill Ventures and cofounded Bracket Computing.

---