Tracy Yang, Juan Alonso, Jason Lango & Tom Economon
Luminary Cloud was founded under so much secrecy that chief executive officer Jason Lango couldn’t immediately tell Tom Economon, the company’s second hire, who his mystery co-founder was. Chief technology officer Juan Alonso, Tom’s Ph.D. advisor at Stanford University, didn’t want to influence his former student’s decision.
That cloak of stealth was also essential. The approximately $10 billion computer-aided engineering computing market was dominated by large, established companies. While the legacy incumbents didn’t have the agility of a startup, they had much larger budgets and access to much greater resources than Luminary Cloud.
“It was clear to both of us that this was going to take a unique combination of cloud infrastructure, computational science, GPU computing, visualization, design, UI, etc.,” says Juan. “Bringing together a world-class multidisciplinary team is just very difficult.”
Their very first employee marked a break from traditional CAE players. Instead of a software developer, they brought on Tracy Yang, head of design. Luminary Cloud’s platform needed not just to be super fast and reliably accurate, but to provide an exceptional user experience.
“Left to its own devices, a typical high-tech company will not make an easy-to-use product. It’s just unnatural for a technology company to think that way,” explains Jason. “But ultimately creating a product that is easy to use is what’s necessary to build a huge company.”
Under the initial name of “Silicon Box”—biologists call simulations in silico experiments—they worked out of a conference room at Sutter Hill Ventures’ headquarters in Palo Alto through the fall of 2019 and beginning of 2020. The move to a real office was pushed back when the pandemic hit in early 2020 and most of Silicon Valley adopted a work from home policy. Luminary Cloud’s team wouldn’t be back together under one roof until September 2021, when they opened an office in Redwood City, California.
Like any startup, the boundaries between different roles were fluid. Jason, for example, was both CEO and I.T. support: provisioning user accounts, setting up websites and resolving computing problems while also overseeing the company in general. Juan and Tom were in charge of developing a technology roadmap that made sense and would clearly differentiate Luminary from other companies in the CAE space. Working with Sutter Hill Ventures’ partners, Jason, Juan and Tom were Luminary Cloud’s first sales team. Working from a standard pitch, they made hundreds of Zoom calls to possible customers as they tried to generate leads and more fully understand their market.
“In those early days my horizons expanded because I was mostly used to computational problems in aerospace companies, but we talked to a lot of potential customers. Like the engineer who wanted to simulate fluid mechanics processes to produce cream cheese and sour cream more efficiently without curdling, like the people who are flushing toilets and making sure golf balls go through,” says Juan. “My whole life has been spent with customers that relish achieving 0.1% accuracy. In reality, much of engineering is about developing products more efficiently, more cost effectively, and bringing them to market faster.”
Talking with potential customers upheld their commitment to prioritizing user experience. “There are millions of mechanical engineers out there, and it was an ‘Aha!’ experience to learn how miserable their lives are due to legacy analysis tools,” adds Jason. The average time required by engineers to learn whether their design iterations would work was weeks. Luminary Cloud’s small team felt sure they could bring that down to hours or even minutes.
From the beginning, Sutter Hill Ventures provided significant startup capital—in total this would reach $115 million—enough to allow the young company to develop and grow its platform over several years. But careful attention was needed to allocate those funds and the company’s resources. “One of the major challenges in a startup is just making sure that you’re not going all over the place. That you’re retaining focus and abiding by the principles that you set off with,” explains Jason.
Key to this was not just the overall vision of the final product, but also how they would move towards that. Jason and Juan knew that they would need to rely on experts from two very different worlds, computational physics and computer science, and that they would need to bridge the gap between the two. This would be easy when the team consisted of half a dozen people, but would become progressively more difficult as it expanded tenfold.
To achieve this, they developed a culture consisting of four principal values. They encouraged staff to be entrepreneurial, taking risks as needed, to be rigorous, paying attention to the details, to speak up, being unafraid to have an opinion, and to take ownership, being responsible and able to make quick decisions.
Early on, the company needed to decide whether to base its computational fluid dynamics (CFD) solver on finite volume or lattice Boltzmann methods, two different approaches for how to discretize and solve the governing equations of fluid flow. At a meeting of the entire team, Jason and Juan invited opinions from everyone. Finite-volume methods were tried and tested and could support advanced capabilities like sensitivity analysis that could help end-users not just leverage the platform for simulation, but for design, too. Lattice-Boltzmann methods had some advantages for transient flows but ignored the fact that many engineering problems are steady.
“Most people had the conventional wisdom that a lattice Boltzmann solver was necessary to get good GPU performance. I think our decision was foundational. It was foundational to Luminary doing things differently,” says Jason. “We zigged when the markets zagged.”
That move was validated when Luminary Cloud’s platform first came online around the time they moved into their new office in Redwood City. “The proof that our approach was working was that in about a year of work with a small, talented team we had the fastest compressible solver in the world,” explains Juan. Compared to code run via on-premises hardware, early demonstrations with prospective customers were showing run times of up to 100 times faster using Luminary Cloud’s GPU-based, cloud-native architecture.
Luminary launched an alpha program with an initial group of a dozen companies in December 2020, and then transitioned to beta 14 months later. The company’s first real customer agreement, with Elroy Air, which is building electric vertical take-off and landing aircraft with the goal of bringing same-day shipping to everyone around the world, was signed on August 31, 2022. Within the first year, the number of customers has steadily expanded to include 37 different companies spanning from eVTOL to airships, water valves to air conditioning, and cycling to golf.
Over the course of the past year, the platform has been further developed and fine-tuned to be production-ready and incorporate automated mesh generation, porous media and moving parts. The user interface has been redesigned into a consumer-ready product, with detailed user documentation, sample projects, powerful visualization tools and controls, and a Python software development kit. Luminary Cloud transitioned to Limited Availability in July 2023 and moved into new offices in San Mateo this past January. The next step in its journey is taking its ground-breaking product out of stealth at NVIDIA’s 2024 GPU Technology Conference.
