A lot of the buzz these days in artificial intelligence is around generative AI and how AI is being used to accelerate software and products for consumers. Today, an AI startup called PhysicsX — co-founded by a former Formula One engineering superstar and a computer science whizz — is emerging from stealth with a very specific focus on building and operating physical systems in the enterprise world.
London-based PhysicsX has come up with an AI platform to create and run simulations for engineers working on project areas like automotive, aerospace and materials science manufacturing — industries where there are regularly bottlenecks in development due to how models are tested before production. It is coming out of stealth today with $32 million in funding.
The round, a Series A, is being led by General Catalyst. Others in the round include a very interesting mix of financial and strategic backers. They include Standard Investments, NGP, Radius Capital and KKR co-founder and co-executive chairman, Henry Kravis. The funding will be used for business development, and to continue developing the company’s platform. This is PhysicsX’s first outside funding.
PhysicX is tackling a problem that has been very consistent yet overlooked in the worlds of manufacturing and physical production.
In any physical system, be it in an experimental lab or a live industrial environment, whenever a new idea is introduced — say, a theory about improving the operating efficiency of a piece of machinery, not to mention work on completely new products — the engineers need to simulate how the new idea will work before committing to developing it, and to further improve how it works. Typically, that simulation and testing work is carried by scientists, engineers who might use some AI in the process but are ultimately working out the process manually.
“Something like airflow across an object may take you an hour or two hours, but if you want to simulate something more complex, it may take you a day or longer. So, there’s a computational cost and therefore also a time cost to this. And that limits the depth at which you can optimize,” said Robin Tuluie, who co-founded PhysicsX with Jacomo Corbo, in an interview.
The pair very much know the pain points firsthand.
Tuluie has already had two different lives as a theoretical physicist. As an academic, he worked alongside Nobel Prize winners with a focus on astrophysics. He then moved into the world of racing, first at Renault and then Mercedes, respectively as head of R&D and chief scientist, where he devised designs that helped his teams win four Formula One world championships (gaining some renown himself in the process). He’s also spent years at Bentley and Volkswagen working on automotive design.
Corbo, who got his PhD from Harvard, has also worked in racing but more recently he founded and headed up QuantumBlack, the AI labs at McKinsey, working with a number of Formula One as well as other automotive and industrial clients on thorny product engineering problems.
The pair have put together a team of no less than 50 scientists — other mechanical engineering specialists, physicists and more — to build out the PhysicsX platform, which is tackling automotive but also a much wider range of applications, said Corbo.
“We are building an enterprise platform to support a pretty broad range of domain applications that are tied to building and optimization problems, physics simulation bottlenecks,” he said. “What PhysicsX buys you is the ability to be able to predict the physics [of a system] with very, very high accuracy and fidelity, doing it, anywhere from 10,000 to a million times faster. Now we can be a whole lot more sophisticated about, for example, mining, across a very high dimensional space.”
PhysicsX’s emergence is coming at a very timely moment in the world of deep learning and AI, specifically in how it is getting applied to the physical world.
It was only earlier this month that DeepMind released new research on how it was applying very advanced machine learning to the world of short- and long-term weather prediction, and Corbo believes that physical turn will underscore the next frontier of AI research and development.
“This is the first time that AI models, these deep learning models, these geometric deep learning models, are overtaking numerical simulation for weather,” Corbo pointed out. “We’re starting to see that happen across physics more broadly. And, that enables a lot of different applications in the space of engineering, which is why we’re building a platform to be able to do that across sectors and across a broad range of domain problems.”
Enterprises have, more generally, hit a lot of snags when it comes to digital transformation — ripping out existing infrastructure to adopt more modern IT and approaches. Although you can classify what PhysicsX is doing as a kind of “digital transformation” too, the startup is able to sidestep those challenges, since the kind of applications it is tackling, in engineering and R&D, are not typically IT issues that require scaling across organizations more widely.
All the same, it’s a new approach, and one that will disrupt how industrial companies approach development today. General Catalyst is therefore both taking a bet on a very hot area — AI — but also breaking some new ground by backing a startup believes how that hot area will evolve.
“PhysicsX expands engineering boundaries in crucial sectors, led by a team deeply skilled in simulation engineering and machine learning,” Larry Bohn, managing director of General Catalyst, said in a statement. “With credibility, customer relationships, and technical expertise, we believe PhysicsX is poised to transform engineering in complex industries. This aligns with our vision for industrial transformation and positions PhysicsX with the opportunity to create a category-defining company in advanced industries.”