Qiqi Wang of FlexCompute, co-founder and associate professor at MIT, discusses how physics-based artificial intelligence AI and graphics processing unit GPU accelerated simulation reshape engineering workflows during theCUBE Research conversation hosted by John Furrier of theCUBE at the NYSE Wired Space Tech event. Wang describes FlexCompute's approach, which combines foundational physics models, automated geometry handling and model-context application programming interfaces API that enable engineers and AI agents to pose quantitative "what if" questions and receive answers at the speed of thought.
Wang emphasizes the need to modernize engineering tools to match software innovation. They highlight that FlexCompute's platform integrates GPU compute, high performance computing HPC, fluid dynamics and other physics disciplines as well as geometry automation, reducing simulation timelines from weeks or months to near real time. Furrier highlights implications for spacecraft docking, defense technology, semiconductors and faster hardware iteration cycles across aerospace and industrial sectors, including digital twin and robotics applications.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Physical AI & Robotics Leaders. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Qiqi Wang, MIT
Qiqi Wang of FlexCompute, co-founder and associate professor at MIT, discusses how physics-based artificial intelligence AI and graphics processing unit GPU accelerated simulation reshape engineering workflows during theCUBE Research conversation hosted by John Furrier of theCUBE at the NYSE Wired Space Tech event. Wang describes FlexCompute's approach, which combines foundational physics models, automated geometry handling and model-context application programming interfaces API that enable engineers and AI agents to pose quantitative "what if" questions and receive answers at the speed of thought.
Wang emphasizes the need to modernize engineering tools to match software innovation. They highlight that FlexCompute's platform integrates GPU compute, high performance computing HPC, fluid dynamics and other physics disciplines as well as geometry automation, reducing simulation timelines from weeks or months to near real time. Furrier highlights implications for spacecraft docking, defense technology, semiconductors and faster hardware iteration cycles across aerospace and industrial sectors, including digital twin and robotics applications.
>> Welcome back everyone. I'm John Furrier, host of theCUBE, here at theCUBE's NYSE studio. Of course, we have our Palo Alto studio connecting Silicon Valley here to Wall Street. This is our Robotics Physical AI Series and Qiqi Wang is here. He's the co-founder of FlexCompute and also associate professor at MIT. Course 16, if everyone's interested. Obviously that's really critical for a lot of the AI infrastructure. Also, space and defense tech. Qiqi, thank you for coming into theCUBE today at the NYSE.
Qiqi Wang
>> Yeah, thank you for having me. It's my pleasure.
John Furrier
>> The Space Tech conference is going on upstairs. Thanks for breaking away for some time. The physics, AI, all kind of rhyme together. Look at all the infrastructure build out we've been covering on our AI Factory Series, is also translating right into robotics, physical AI. Defense tech is a big part of that. And you see drones, you see physical AI robotics in the field. No better case study than to solve the hardest problems in space. You have a zillion PhDs on your staff. I looked at the site, not a zillion, but of many. You guys are pioneering in that. Explain what you guys do. Then I'll get into some of the questions.
Qiqi Wang
>> Okay. FlexCompute, as we said, is a physics intelligence company. And we provide the technology for engineers and now increasingly AI agents to ask physics questions. And this is the kind of questions like what if questions that really underpins all human ingenuity. So physically what that means is that we expose the APIs and now increasingly model context protocols, so that engineers and the agents can really ask the quantitative physical questions and get answers at the speed of thought instead of having to set up clunky simulations to get results. And you can think of it as providing a collaborative physical brain so that people can work with to design their product and figure out the hardware solutions that nobody-
John Furrier
>> So it's really assisting the human.
Qiqi Wang
>> Yes.
John Furrier
>> That's not replacing or using... I mean, people lean on AI a lot, but the power users of AI are the ones that drive it.
Qiqi Wang
>> Yes. It's really enabling the human engineers. And basically we build the product as we wanted to use it ourselves.
John Furrier
>> By engineers, for engineers.
Qiqi Wang
>> Exactly.
John Furrier
>> All right. So talk about the origination story. Did it come to you? How'd this idea come to you? And obviously we're seeing AI move out of language to vision. I mean, I see biology models. You got physics, natural progression. Where'd this idea come from? Tell me the origination story.
Qiqi Wang
>> Well, the originating story is a little bit more mundane because we looked at how engineering software has been stagnating for decades. Consumer software has been evolving revolutionarily almost every few years over the past few decades. But engineers has been using pretty much the same tools since like 1990s.
John Furrier
>> Fortran. I won't even bring that name up. Remember the equipment was all run on Fortran code. No, but it's antiquated.
Qiqi Wang
>> Yeah, it's antiquated. And we as engineers ourselves, as I said, I'm a professor in AeroAstro, feel the pain firsthand. So the funding team of FlexCompute are really engineers like me. We have a lot of expertise in high performance computing, but we are also spending a lot of wide range of physics and engineering disciplines. And we came together to really build the tools that we and our colleagues wanted to use. So since the beginning, basically the idea or the model is to use advanced computing to accelerate innovation and specifically to make innovation hardware to be as fast and as easy as innovation in software.
John Furrier
>> And robotics is a great example of some of the innovation speed. I mean, the velocity of the innovation in robotics has been off the charts. We've been seeing great progress, mainly because one, the compute's available. You got hardware, which is grounded in software.
Qiqi Wang
>> Yep.
John Furrier
>> We see that with NVIDIA. Everyone talks about CUDA and the relationship they have, but that's applying everywhere. Talk about that dynamic of the software relationship to hardware and how that's impacting the engineers. Because engineers look at hardware and software as one thing, as most do, but you have specialists that do hardware, specialists that do software. But talk about that blend between software and hardware and how it relates to FlexCompute.
Qiqi Wang
>> Well, software engineering has been very easy in some sense, right? Because every time you change the code, you hit compile and the code actually runs. And if there is any bug, you see it immediately. And if it behaves in a way you don't expect, you also see it immediately. Hardware is different in the sense that when you change your design, you either have to physically build it and test it, or you have to test it in a computational simulation. Either way, it takes tremendous amount of time. And there is a lot of things involved, not just the physics, but the geometry. When you change your design, it usually involves changing geometry and there is a lot of friction in the process. And that is one of the things we are changing as I can give you some examples later on.
John Furrier
>> Yeah. So let's talk simulation. I want to get the example. Let's talk about, you mentioned simulation. Simulation has been a great advantage for using synthetic data, looking at digital twins. But when you look at physics, you have scenarios of actual properties. How has simulation changed and what do you guys do differently?
Qiqi Wang
>> Yes. So simulations has been a great enabler, right? Basically enables people to not really have to build the physical things and test it. And space is a great example. Building things and testing space is extremely slow and you have to wait for it to launch and things like that. So simulation for space is especially important, but simulation tools has been really stagnant for a while. And that's because all the simulations has been built on CPU platforms. And in order to even run one simulation of one design, one operating condition, engineers has often has to wait for hours, sometimes even days for that simulation to finish. So that became really the pacing item for hardware innovation, right? The bottleneck for hardware innovation has not been the vision of what could be built. It was not the ambition for building things that people envision could be built. It's not engineer's ingenuity. It's the tool. It's the tool that has been throttling how much we advance in hardware engineering.
John Furrier
>> Yeah. It's limiting at big time. One of the things about space, since you mentioned this focus in your focus, is that it's hard to unwind that. You launch something into space, it's up there.
Qiqi Wang
>> Yes.
John Furrier
>> If it runs out of battery and fuel, it's over. And if it breaks, you can't send the technician up. You can't upload. You can upload, I guess now, software, but the hardware has to be tested.
Qiqi Wang
>> Yes.
John Furrier
>> Talk about that requirement. That's pretty stringent. That's a constraint. How do you guys look at that? And give us some examples of how you guys are working with things like that.
Qiqi Wang
>> Yeah, that's a great question. So I'm going to give an example of us working with Northrop Grumman. And an example is with spacecraft docking. And for people who are not familiar, spacecraft docking, as you said, is an extremely high consequence and also incredibly dangerous maneuver. And this is because you have to fire thrusters, rocket thrusters to position yourself to dock and also to decelerate. And these thrusters can easily push your target into spinning. And if your target is spinning, you cannot dock. It can also literally melt and damage through the hot gas from the thrusters' different components. And also you may risk depositing unburned propellants into sensitive equipments that can also terminate the mission. So this is an extremely critical application in which engineers absolutely have to prioritize the accuracy of the simulations of how the plumes of your thruster behave in the vacuum of space. And the Northrop Grumman engineers 100% prioritize the accuracy over anything else. And that's why each simulation takes weeks. And in order to plan one single docking maneuver, it takes months of simulations to do that planning. And we completely changed that. Our AI model, our physics AI model enables such planning to now be performed in real time rather than over months.
John Furrier
>> So two things happening. One is it's the compute was slow. I mean, I talked to someone at Boeing once on theCUBE, we did an interview. And doing wing design, just all the variations of what could happen on an airplane.
Qiqi Wang
>> Yes.
John Furrier
>> It's too complex. It would be weeks before they get their models back. So GPUs come. Thank you very much. Check. Now the AI software. Talk more about that AI software advantage. Because now you have the perfect storm on the tooling because you got the GPUs and you got the AI. Talk about that piece.
Qiqi Wang
>> Yes. So the coupling between the GPU computing and AI is critical. Actually, there is also a third piece that's the geometry handling. So FlexCompute built our infrastructure around the three pillars. One is the GPU computing. The GPU computing, we have a simulation of foundational physics, including fluid dynamics. Fluid dynamics covers the rarefied gas that's happening in space. We also have structure. We also have electromagnetics. We also have thermal. That's one of the most important part. And the second part is actually geometry. So we use AI to automatically handle geometry that resolves one of the most important bottleneck of simulation. That's the manual processing of a lot of the geometry. And these two components really enables the AI part that makes it possible to train AI that engineers can trust in this kind of critical applications like spacecraft docking.
John Furrier
>> You guys have a lot of customers. You mentioned one, the docking one that's very complex. Talk about some of the other customers that you're engaging with. Share some of the use cases in the physics world and space. What else are you working on? You got a wide range. I mean, physics is everywhere. I mean, the laws of physics, I've heard that quote many times, speed of light, laws of physics. You can't really change physics, but you can understand them.
Qiqi Wang
>> Yes. Yeah. So our customers really covers a big range. So in aerospace and the defense, we have Northrop Grumman, we have Embraer, we have Joby Aviation who is being listed as of today. And yeah, we have a lot of other people who are doing defense. And we have JetZero. But apart from that, we also have a lot of customers in semiconductors. We have Samsung. We have Global Foundry. We have Xanadu. We have basically a lot of the people who are not just designing the most innovative aircraft and spacecraft, but also pushing the limits of both silicon and quantum computing.
John Furrier
>> So basically anything with material science involved, you can play.
Qiqi Wang
>> Yes.
John Furrier
>> Because chips, they're not in space, but they will be in space.
Qiqi Wang
>> Yeah.
John Furrier
>> Space is natural because you have the tooling. So it comes down to the tools. You're building the tooling layer to be very compatible with engineers.
Qiqi Wang
>> Yes. I mean, physics is physics, right? Basically we provide the physics layer. And wherever that particular physical disciplines from the application engineering, we have a customer.
John Furrier
>> All right. What's next? Tell us what's next. And how old is the company?
Qiqi Wang
>> Well, what's next is I talked about the three pillars, right? And the first pillar is a foundational simulation. And we are just going to be deepening that layer by involving more and more physical disciplines. And we are also going to be improving our geometry AI capabilities. And that is going to allow us to expand into more engineering applications that has high consequences. And our principle is to train physical AI that doesn't really just approximate the behavior of these physical laws, but really train them to understand the physical laws themselves. And that's kind of the difference between AI models that works just most of the time versus AI models that engineers can trust.
John Furrier
>> You're optimizing for accuracy because of the criticality nature of what you do for a 100% accuracy.
Qiqi Wang
>> Yes. We are optimizing for trustworthiness, right? For example, in the spacecraft docking application, we are not training AI models to approximate the behavior of the plume of the rocket, of how it reflects off a spacecraft, how it forms a shock layer. No. We are training the AI to understand how one molecule interacts with other molecules in such interactions, right? So this is a philosophical difference of what the AI actually does, right?
John Furrier
>> Yeah.
Qiqi Wang
>> And that makes the difference between AI that works.
John Furrier
>> You're rooted in engineering principles.
Qiqi Wang
>> Exactly.
John Furrier
>> And you unpack the key elements and train it on that with the geometry piece.
Qiqi Wang
>> Exactly. Exactly.
John Furrier
>> Okay. So how old is the company?
Qiqi Wang
>> The company is a little bit more than a decade old.
John Furrier
>> Okay. So you got a lot of PhDs, you got a lot of staff, and train it on that with the geometry piece.
Qiqi Wang
>> Exactly. Exactly.
John Furrier
>> Okay. So how old is the company?
Qiqi Wang
>> The company is a little bit more than a decade old.
John Furrier
>> Okay. So you've got a lot of PhDs, you've got a lot of staff. How's it going at MIT? Good recruiting there? A lot of brainiacs go to MIT?
Qiqi Wang
>> Yes, very good.
John Furrier
>> You get a lot of PhDs.
Qiqi Wang
>> Yes.
John Furrier
>> Talk about the MIT angle here. Actually, you teach a class there, a lot of smart people at MIT, a lot of PhDs. What's it like there? A lot of PhD students you're recruiting there. What are some of the things that... What's the culture like?
Qiqi Wang
>> Yeah. So this is very interesting because at MIT, I see a lot of great students and I can also collaborate with all kinds of professors. The whole MIT is, I can say, a huge physics department in the sense that people, even in different engineering departments, works on different physics, right? And we can work with a lot of different professors. We can recruit from different groups into FlexCompute. And inside of FlexCompute, the culture is very much academic, right? People collaborate with each other, but with a key difference of working together on problems with significant engineering consequences. So that's what I found the most fascinating about the culture of FlexCompute.
John Furrier
>> Yeah. PhDs like to solve hard problems and you got one of them. Final question for you. Upstairs at the Space Conference here at the NYSE, what's some of the conversations like inside the building today? What are some of your peers and what's the industry's top issues being worked on?
Qiqi Wang
>> Oh, there is a lot of great issues. And one of the most fascinating thing I found the conversation upstairs is I came over, this is NYSE, and I was expecting to talking to executives and the business people. But what I found is a great mixture of a lot of leaders who are very, very deeply technical and we understand each other immediately. And this is a great sign for the future, right? And what I expect from today's conversation is that the next few decades is going to be as innovative as software for the past few decades. So back in the 1980s and the '90s, nobody, I say nobody can predict the kind of a software ecosystem we're using right now. Nobody can predict the internet we are surfing right now. Nobody can predict the array of gadgets we carry with us today, right? I would say the next few decades for space is just going to be like that. And I expect that today, none of us can expect, can predict what we have in 20, 30, 40 years from now. And the FlexCompute is building the physical intelligence layer that's making that possible.
John Furrier
>> Well, you're a frontier leader and it's going to be all unexpected. The future's unwritten and you guys are doing a great job. Thanks for coming on theCUBE and we're doing our part here. NYSE's got a great community. A lot of technical people, not just finance and business, got a great community. Of course, the NYSE Wired, theCUBE's brand program is doing our part. Thanks for coming on theCUBE. Appreciate your time.
Qiqi Wang
>> Yeah, thank you for having me. It's a great pleasure speaking with you.
John Furrier
>> Okay. I'm John Furrier, host of theCUBE here at theCUBE's NYSE Wired program. An open community, of course, NYSE bringing the leaders together, the frontier leaders. Thanks for watching.
>> Welcome back everyone. I'm John Furrier, host of theCUBE, here at theCUBE's NYSE studio. Of course, we have our Palo Alto studio connecting Silicon Valley here to Wall Street. This is our Robotics Physical AI Series and Qiqi Wang is here. He's the co-founder of FlexCompute and also associate professor at MIT. Course 16, if everyone's interested. Obviously that's really critical for a lot of the AI infrastructure. Also, space and defense tech. Qiqi, thank you for coming into theCUBE today at the NYSE.
Qiqi Wang
>> Yeah, thank you for having me. It's my pleasure.
John Furrier
>> The Space Tech conference is going on upstairs. Thanks for breaking away for some time. The physics, AI, all kind of rhyme together. Look at all the infrastructure build out we've been covering on our AI Factory Series, is also translating right into robotics, physical AI. Defense tech is a big part of that. And you see drones, you see physical AI robotics in the field. No better case study than to solve the hardest problems in space. You have a zillion PhDs on your staff. I looked at the site, not a zillion, but of many. You guys are pioneering in that. Explain what you guys do. Then I'll get into some of the questions.
Qiqi Wang
>> Okay. FlexCompute, as we said, is a physics intelligence company. And we provide the technology for engineers and now increasingly AI agents to ask physics questions. And this is the kind of questions like what if questions that really underpins all human ingenuity. So physically what that means is that we expose the APIs and now increasingly model context protocols, so that engineers and the agents can really ask the quantitative physical questions and get answers at the speed of thought instead of having to set up clunky simulations to get results. And you can think of it as providing a collaborative physical brain so that people can work with to design their product and figure out the hardware solutions that nobody-
John Furrier
>> So it's really assisting the human.
Qiqi Wang
>> Yes.
John Furrier
>> That's not replacing or using... I mean, people lean on AI a lot, but the power users of AI are the ones that drive it.
Qiqi Wang
>> Yes. It's really enabling the human engineers. And basically we build the product as we wanted to use it ourselves.
John Furrier
>> By engineers, for engineers.
Qiqi Wang
>> Exactly.
John Furrier
>> All right. So talk about the origination story. Did it come to you? How'd this idea come to you? And obviously we're seeing AI move out of language to vision. I mean, I see biology models. You got physics, natural progression. Where'd this idea come from? Tell me the origination story.
Qiqi Wang
>> Well, the originating story is a little bit more mundane because we looked at how engineering software has been stagnating for decades. Consumer software has been evolving revolutionarily almost every few years over the past few decades. But engineers has been using pretty much the same tools since like 1990s.
John Furrier
>> Fortran. I won't even bring that name up. Remember the equipment was all run on Fortran code. No, but it's antiquated.
Qiqi Wang
>> Yeah, it's antiquated. And we as engineers ourselves, as I said, I'm a professor in AeroAstro, feel the pain firsthand. So the funding team of FlexCompute are really engineers like me. We have a lot of expertise in high performance computing, but we are also spending a lot of wide range of physics and engineering disciplines. And we came together to really build the tools that we and our colleagues wanted to use. So since the beginning, basically the idea or the model is to use advanced computing to accelerate innovation and specifically to make innovation hardware to be as fast and as easy as innovation in software.
John Furrier
>> And robotics is a great example of some of the innovation speed. I mean, the velocity of the innovation in robotics has been off the charts. We've been seeing great progress, mainly because one, the compute's available. You got hardware, which is grounded in software.
Qiqi Wang
>> Yep.
John Furrier
>> We see that with NVIDIA. Everyone talks about CUDA and the relationship they have, but that's applying everywhere. Talk about that dynamic of the software relationship to hardware and how that's impacting the engineers. Because engineers look at hardware and software as one thing, as most do, but you have specialists that do hardware, specialists that do software. But talk about that blend between software and hardware and how it relates to FlexCompute.
Qiqi Wang
>> Well, software engineering has been very easy in some sense, right? Because every time you change the code, you hit compile and the code actually runs. And if there is any bug, you see it immediately. And if it behaves in a way you don't expect, you also see it immediately. Hardware is different in the sense that when you change your design, you either have to physically build it and test it, or you have to test it in a computational simulation. Either way, it takes tremendous amount of time. And there is a lot of things involved, not just the physics, but the geometry. When you change your design, it usually involves changing geometry and there is a lot of friction in the process. And that is one of the things we are changing as I can give you some examples later on.
John Furrier
>> Yeah. So let's talk simulation. I want to get the example. Let's talk about, you mentioned simulation. Simulation has been a great advantage for using synthetic data, looking at digital twins. But when you look at physics, you have scenarios of actual properties. How has simulation changed and what do you guys do differently?
Qiqi Wang
>> Yes. So simulations has been a great enabler, right? Basically enables people to not really have to build the physical things and test it. And space is a great example. Building things and testing space is extremely slow and you have to wait for it to launch and things like that. So simulation for space is especially important, but simulation tools has been really stagnant for a while. And that's because all the simulations has been built on CPU platforms. And in order to even run one simulation of one design, one operating condition, engineers has often has to wait for hours, sometimes even days for that simulation to finish. So that became really the pacing item for hardware innovation, right? The bottleneck for hardware innovation has not been the vision of what could be built. It was not the ambition for building things that people envision could be built. It's not engineer's ingenuity. It's the tool. It's the tool that has been throttling how much we advance in hardware engineering.
John Furrier
>> Yeah. It's limiting at big time. One of the things about space, since you mentioned this focus in your focus, is that it's hard to unwind that. You launch something into space, it's up there.
Qiqi Wang
>> Yes.
John Furrier
>> If it runs out of battery and fuel, it's over. And if it breaks, you can't send the technician up. You can't upload. You can upload, I guess now, software, but the hardware has to be tested.
Qiqi Wang
>> Yes.
John Furrier
>> Talk about that requirement. That's pretty stringent. That's a constraint. How do you guys look at that? And give us some examples of how you guys are working with things like that.
Qiqi Wang
>> Yeah, that's a great question. So I'm going to give an example of us working with Northrop Grumman. And an example is with spacecraft docking. And for people who are not familiar, spacecraft docking, as you said, is an extremely high consequence and also incredibly dangerous maneuver. And this is because you have to fire thrusters, rocket thrusters to position yourself to dock and also to decelerate. And these thrusters can easily push your target into spinning. And if your target is spinning, you cannot dock. It can also literally melt and damage through the hot gas from the thrusters' different components. And also you may risk depositing unburned propellants into sensitive equipments that can also terminate the mission. So this is an extremely critical application in which engineers absolutely have to prioritize the accuracy of the simulations of how the plumes of your thruster behave in the vacuum of space. And the Northrop Grumman engineers 100% prioritize the accuracy over anything else. And that's why each simulation takes weeks. And in order to plan one single docking maneuver, it takes months of simulations to do that planning. And we completely changed that. Our AI model, our physics AI model enables such planning to now be performed in real time rather than over months.
John Furrier
>> So two things happening. One is it's the compute was slow. I mean, I talked to someone at Boeing once on theCUBE, we did an interview. And doing wing design, just all the variations of what could happen on an airplane.
Qiqi Wang
>> Yes.
John Furrier
>> It's too complex. It would be weeks before they get their models back. So GPUs come. Thank you very much. Check. Now the AI software. Talk more about that AI software advantage. Because now you have the perfect storm on the tooling because you got the GPUs and you got the AI. Talk about that piece.
Qiqi Wang
>> Yes. So the coupling between the GPU computing and AI is critical. Actually, there is also a third piece that's the geometry handling. So FlexCompute built our infrastructure around the three pillars. One is the GPU computing. The GPU computing, we have a simulation of foundational physics, including fluid dynamics. Fluid dynamics covers the rarefied gas that's happening in space. We also have structure. We also have electromagnetics. We also have thermal. That's one of the most important part. And the second part is actually geometry. So we use AI to automatically handle geometry that resolves one of the most important bottleneck of simulation. That's the manual processing of a lot of the geometry. And these two components really enables the AI part that makes it possible to train AI that engineers can trust in this kind of critical applications like spacecraft docking.
John Furrier
>> You guys have a lot of customers. You mentioned one, the docking one that's very complex. Talk about some of the other customers that you're engaging with. Share some of the use cases in the physics world and space. What else are you working on? You got a wide range. I mean, physics is everywhere. I mean, the laws of physics, I've heard that quote many times, speed of light, laws of physics. You can't really change physics, but you can understand them.
Qiqi Wang
>> Yes. Yeah. So our customers really covers a big range. So in aerospace and the defense, we have Northrop Grumman, we have Embraer, we have Joby Aviation who is being listed as of today. And yeah, we have a lot of other people who are doing defense. And we have JetZero. But apart from that, we also have a lot of customers in semiconductors. We have Samsung. We have Global Foundry. We have Xanadu. We have basically a lot of the people who are not just designing the most innovative aircraft and spacecraft, but also pushing the limits of both silicon and quantum computing.
John Furrier
>> So basically anything with material science involved, you can play.
Qiqi Wang
>> Yes.
John Furrier
>> Because chips, they're not in space, but they will be in space.
Qiqi Wang
>> Yeah.
John Furrier
>> Space is natural because you have the tooling. So it comes down to the tools. You're building the tooling layer to be very compatible with engineers.
Qiqi Wang
>> Yes. I mean, physics is physics, right? Basically we provide the physics layer. And wherever that particular physical disciplines from the application engineering, we have a customer.
John Furrier
>> All right. What's next? Tell us what's next. And how old is the company?
Qiqi Wang
>> Well, what's next is I talked about the three pillars, right? And the first pillar is a foundational simulation. And we are just going to be deepening that layer by involving more and more physical disciplines. And we are also going to be improving our geometry AI capabilities. And that is going to allow us to expand into more engineering applications that has high consequences. And our principle is to train physical AI that doesn't really just approximate the behavior of these physical laws, but really train them to understand the physical laws themselves. And that's kind of the difference between AI models that works just most of the time versus AI models that engineers can trust.
John Furrier
>> You're optimizing for accuracy because of the criticality nature of what you do for a 100% accuracy.
Qiqi Wang
>> Yes. We are optimizing for trustworthiness, right? For example, in the spacecraft docking application, we are not training AI models to approximate the behavior of the plume of the rocket, of how it reflects off a spacecraft, how it forms a shock layer. No. We are training the AI to understand how one molecule interacts with other molecules in such interactions, right? So this is a philosophical difference of what the AI actually does, right?
John Furrier
>> Yeah.
Qiqi Wang
>> And that makes the difference between AI that works.
John Furrier
>> You're rooted in engineering principles.
Qiqi Wang
>> Exactly.
John Furrier
>> And you unpack the key elements and train it on that with the geometry piece.
Qiqi Wang
>> Exactly. Exactly.
John Furrier
>> Okay. So how old is the company?
Qiqi Wang
>> The company is a little bit more than a decade old.
John Furrier
>> Okay. So you got a lot of PhDs, you got a lot of staff, and train it on that with the geometry piece.
Qiqi Wang
>> Exactly. Exactly.
John Furrier
>> Okay. So how old is the company?
Qiqi Wang
>> The company is a little bit more than a decade old.
John Furrier
>> Okay. So you've got a lot of PhDs, you've got a lot of staff. How's it going at MIT? Good recruiting there? A lot of brainiacs go to MIT?
Qiqi Wang
>> Yes, very good.
John Furrier
>> You get a lot of PhDs.
Qiqi Wang
>> Yes.
John Furrier
>> Talk about the MIT angle here. Actually, you teach a class there, a lot of smart people at MIT, a lot of PhDs. What's it like there? A lot of PhD students you're recruiting there. What are some of the things that... What's the culture like?
Qiqi Wang
>> Yeah. So this is very interesting because at MIT, I see a lot of great students and I can also collaborate with all kinds of professors. The whole MIT is, I can say, a huge physics department in the sense that people, even in different engineering departments, works on different physics, right? And we can work with a lot of different professors. We can recruit from different groups into FlexCompute. And inside of FlexCompute, the culture is very much academic, right? People collaborate with each other, but with a key difference of working together on problems with significant engineering consequences. So that's what I found the most fascinating about the culture of FlexCompute.
John Furrier
>> Yeah. PhDs like to solve hard problems and you got one of them. Final question for you. Upstairs at the Space Conference here at the NYSE, what's some of the conversations like inside the building today? What are some of your peers and what's the industry's top issues being worked on?
Qiqi Wang
>> Oh, there is a lot of great issues. And one of the most fascinating thing I found the conversation upstairs is I came over, this is NYSE, and I was expecting to talking to executives and the business people. But what I found is a great mixture of a lot of leaders who are very, very deeply technical and we understand each other immediately. And this is a great sign for the future, right? And what I expect from today's conversation is that the next few decades is going to be as innovative as software for the past few decades. So back in the 1980s and the '90s, nobody, I say nobody can predict the kind of a software ecosystem we're using right now. Nobody can predict the internet we are surfing right now. Nobody can predict the array of gadgets we carry with us today, right? I would say the next few decades for space is just going to be like that. And I expect that today, none of us can expect, can predict what we have in 20, 30, 40 years from now. And the FlexCompute is building the physical intelligence layer that's making that possible.
John Furrier
>> Well, you're a frontier leader and it's going to be all unexpected. The future's unwritten and you guys are doing a great job. Thanks for coming on theCUBE and we're doing our part here. NYSE's got a great community. A lot of technical people, not just finance and business, got a great community. Of course, the NYSE Wired, theCUBE's brand program is doing our part. Thanks for coming on theCUBE. Appreciate your time.
Qiqi Wang
>> Yeah, thank you for having me. It's a great pleasure speaking with you.
John Furrier
>> Okay. I'm John Furrier, host of theCUBE here at theCUBE's NYSE Wired program. An open community, of course, NYSE bringing the leaders together, the frontier leaders. Thanks for watching.
