Podcast: Quantum-Accelerated Digital Twins for Aerospace & Defense

Introduction to the Podcast

Host: Woodrow Bellamy: Welcome back to season 10 of the Aerospace and Defense Technology Podcast. My name is Woodrow Bellamy III. I'm a Senior Editor with SAE Media Group, and this season of the podcast is all about the future of military computing.

On our first episode of the season, we talked about a new in-memory computing chip for on-device artificial intelligence applications, and on this episode, we'll be discussing another future-facing topic with a focus on quantum technology. BQP is a quantum-first simulation startup that has established partnerships and financial backing with the U.S. Air Force Research Lab, IBM, Intel, and Moog, among others. The Syracuse, New York-based startup is developing what it describes as a next-generation simulation backbone for aerospace defense and semiconductor designers.

They have a focus on eliminating the speed, accuracy, and computational performance limitations of legacy simulation algorithms that are currently used by designers to develop new aerospace and defense technologies, such as aircraft, drones, ground vehicles, and other platforms. Abhishek Chopra is the CEO and founder of BQP, and he is the guest on this episode of the podcast to explain how their simulation platform is using quantum algorithms on commercially available high-performance computing technology today. So, here's our interview with Abhishek Chopra, CEO and founder of BQP.

BQP Overview and Background

WB: So, Abhishek, can you start with a brief introduction to who BQP is and some background on the type of technology you're providing for the aerospace and defense industry?

Abhishek Chopra: Absolutely. So, BQP is a dual-use startup in modeling and simulation space. We are solving the bottleneck of time and accuracy that the engineers have to suffer with quantum algorithms on existing hardware today, while also getting ready for when quantum computers become industrial scale.

And specifically for the aerospace and defense teams, we are helping them to basically supercharge their existing simulation workflows with these solvers that we are building on top of the quantum algorithms.

Company Location and Founding Story

WB: And just maybe a follow-up question on that. Can you tell us a little bit about where you're located and also just kind of how long you guys have been doing this as well?

AC: Sure. We are actually headquartered in Syracuse in New York. We have been an upstate New York company. We are actually origins from India and then we moved, like myself and my co-founder were here.

So, we then established the US entity, which is when we came full-time. We got our first funding. But primarily, to give you a sense of the founding story of why BQP and why everything was started.

So, myself, I'm an aerospace engineer and a computational scientist by trade. I was developing these large-scale modeling simulation tools, particularly in computational fluid dynamics for aerospace problems like electric vehicles, wind turbines, and aircrafts. My biggest simulation actually took six months on 250,000 computers running parallelly.

So, I asked myself this question that this is not going to be useful because in engineering, we don't test a component one time. We tend to test it like 10,000 times. So, if six months into 10,000, it's a lot of time and it's just infeasible for industry today.

So, I wanted to solve this problem of slow and expensive simulation. I worked with Oak Ridge National Lab and NVIDIA to basically bring this legacy Fortran C code from CPUs to GPUs and got the opportunity to even run on one of the largest supercomputers of the world at that time. But we got suboptimal performance improvement.

The reason for that is because the map that goes behind these tools, basically the algorithms, data structures, and so on and so forth, is just outdated and haven't changed in past 40 years. So, that led to basically the reason to start BQP. I started this with my best buddy from undergrad, Ruth Lineswala, who actually went to University of Minnesota working in hypersonic simulation.

So, he was facing exactly the same issues, just in a different use case within the industry. My third co-founder who was facing similar issue in finance and quantitative finance with their kind of simulations and partial differential equations. So, we all joined hand.

This was back in 2020. Yeah, that's the founding story for BQP.

Understanding Quantum Computing and BQP’s Approach

WB: Yeah, it's a very interesting and inspiring story.

And it's really interesting to see the sort of announcement that recently came out, some of the seed round funding that you won. And taking a look at your website, it notes that you are a simulation company, as you mentioned, that leverages quantum technology for high-speed computing and future computing. So, this second question, I just want to do kind of a two-part question.

So, you know, quantum computing, quantum technology, we've covered it a lot. A lot of people are familiar with it, but I think it's always good to start out with kind of a definition of what quantum computing is when we discuss it. And then the second part of that question would be, what elements of quantum are you leveraging in your simulation product today, as it stands today? So, first part of the question, just give us a basic overview of what quantum computing is.

AC: Yeah. So, if it is okay, I'll try to switch because then it will become very clear on what we at BQP do. So, whenever we hear the word quantum, the first thing that comes to most of the people in mind is quantum computers.

And then immediately the thought comes, oh, it's a technology of future, it's 10 years away. What is not very well known is the world of quantum information science actually preceded the field of quantum computing. And the math was laid out there.

It is the math that was seen and was said that, oh, wait a minute, this can be done better on a quantum mechanical system, which is what the birth of quantum computing field was. And there has been two revolutions around it. The first was in early 80s and 90s when people started to think about making quantum computer.

And now, when we are starting to become industrial scale and more broader use for quantum computer. But it's same way I say, but that doesn't mean that the math cannot run on CPUs and GPUs. It just runs better on a quantum mechanical system.

Same way as saying AI runs better on a GPU doesn't mean that it cannot run on CPUs. Now, comes the quantum computing. So, we just talked about quantum computing has the math part to it and then putting the math onto the quantum mechanical system.

What we are doing today is the math part of it. We are leveraging stochastic linear algebra, probabilistic linear algebra in certain of our solvers like optimization, which has lend itself to doing large scale nonlinear, non-convex type of optimization problem, which is almost everywhere in aerospace defense and some other industries which are critical. And that's what we are leveraging.

We are leveraging that on GPUs today. Now, one thing to note here is, as I mentioned, the math from my anecdote earlier, because of the math, you are utilizing maybe 30% of your GPU compute for engineering simulation I'm talking about. By using these better algorithms, we are able to go closer to realizing the full potential of GPUs.

Now, at a point of time, we hear this quite often today, especially because of AI, that the compute is going to become the bottleneck. So, after you have got 100% of every GPU in the world, what after that? You still will not be able to solve the big problems in aerospace defense, including doing analysis around a full aircraft, which is said to be not possible until 2070. Let's say more on the defense side, drone swarm, wargaming scenario modeling.

These are things that you would ideally like to do real time, especially like digital twin and model-based system engineering. And that's where you need more compute. This is another myth that I would like to break, that the future is not just going to be quantum.

The future is, it's going to be CPU, GPU and quantum computer all sitting next to each other. And your problem basically you have to decipher as what we are doing is a vectorization problem, which part of your calculation has to go on which computer.

Work with Defense and Aerospace Organizations

WB: Got it.

WB: Okay. That makes a lot of sense. And thanks for the overview and breakdown, that does make a lot of sense in terms of what you're providing today.

So, as a follow-up to that, we do understand your company has been working with DARPA, Defense Advanced Research Projects Agency, as well as the Air Force Research Lab and several other aerospace and defense companies. What type of work are you doing with them today? And maybe, what are they evaluating your platform for?

AC: So, when we started this and coming both myself and my other co-founder Ruth, we both knew the kind of problems that existed in this world of aerospace defense. So, we first validated that, yes, I mean, you can only do so much with today's modeling and simulation tool.

And there's a bottleneck. You can bring the best of the best GPU, the hardware is not the problem, the map is the problem, as I said. So, when we talked to all of these customers, there's some very, very high-value problems that everybody would like to solve, which comes to digital twin, comes to fluid dynamics, fluid dynamics with structure thermal inputs and mission inputs and all those kinds of things.

And that's really physics-based problems. Then you have the today's era problem, which is machine learning and people using AI machine learning, not just generative AI, but also predictive AI, which is very critical for aerospace industry. And then comes the third layer, which is the optimization.

So, there are tons of optimization problem. And as I said in my previous example, there are some things that can benefit just from quantum math on existing hardware, and that's the optimization. And we are leveraging this quantum-inspired optimization for the names that you mentioned, where there are some very critical problems, whether you call it from a design point of view, or you call it from system design point of view or mission design point of view.

So, with AFRL, for example, we are looking at mission system design, where you have a large-scale system design problem and with mission inputs in it. It's a problem not just for AFRL, but many places as well. And that's what we basically are starting from.

From the commercial aerospace companies, we have looked at design, we have looked at some of the mission problems, like payload type of optimization problems, and they have actually, in the case of AFRL, in case of the commercial aerospace companies, what they're seeing is their input is not changing, their output is not changing, their infrastructure today is not changing, but they are able to do those kinds of problems that they didn't think that they would be able to do before. And that basically goes with the ROI, that few percent improvement in the results by reaching a more optimal point, which the math didn't allow before, and the new math allows you to, three to five percent, one order of magnitude, creates billion dollars of savings and doing intractable problem that were not possible before. And this is nothing to do with quantum computer right now.

Think of what quantum can then present in future for the other problems I mentioned. Yeah, that's a great point. Yeah, like I said, it's a really interesting story and sort of interesting problem you're trying to solve today with leveraging quantum.

Future Development and Vision for BQP

WB: So in terms of next steps, what is BQP focused on in terms of developing, just from infrastructure or compute, or what are you investing in researching and developing from a product standpoint to build your product out and be able to offer this at wide scale and help all these engineers solve these complex design and simulation problems?

AC: Absolutely. So the vision that we are working towards is being the BlackRock of the simulation world. What it means is basically we are not developing the next big modeling simulation software.

No, no, that's not the game we are playing. We are playing the game of building the new novel backend solvers, which will integrate into existing workflows. As I mentioned, the input remains the same, the output remains the same, the infrastructure today remains the same, and I'll come to what we are doing from an infrastructure point of view in a minute.

But the thought process of what's next for us is with this optimization solver that we have, we are currently going after commercial aerospace defense companies. Some of them already are design partners. We have done some paid pilots, and the next step is, can we integrate into their workflows? And that's what we're really testing.

Not only integrating into their workflows, but also integrating with partners, established incumbents in the market, MathWorks, Ansys, Dassault, Siemens, and all, so that we basically, the users don't have to learn any new solver. It's just that they can do 10 times more simulation in an hour than they were able to do before. That's really the value proposition.

In terms of what's next for these customers, so that's where we are going both deeper. We have invested in going deeper with the commercial aerospace, and not just the big logos and the primes you're talking about, but we have seen quite some traction with space companies. And broadly speaking, the defense side, the DoD side is where we are very, very interested in, and actually even DoD is very interested in.

So currently, I'm speaking from Colorado right now, where we just finished our program with the SDA, TAP LAB, Space Domain Awareness under the Space Force. And people are excited, not just about our one solver, but our future solvers also, data-driven solver, machine learning, quantum machine learning, and quantum CFD, for example. And that's the last point here on where we are investing.

Yes, we have found something, an intelligent math to be able to do today, but that doesn't solve the long-term problem. That's what we're investing in, the R&D of quantum native solvers with the problems that the engineering world face. So as I mentioned, predictive machine learning, we are looking into computational fluid dynamics, which is a big theme, which is, again, the background that myself and Ruth come from, and a lot of the team that we have built right now around that.

We've got top experts in modeling simulation, fluid dynamics, high-performance computing, and quantum computing working all together in the same room, essentially, to be able to achieve that vision, the vision of you can do full aircraft analysis, not in 2070, but in the next 10 years. We are crazy enough to think that we can do so. That's what we're working towards.

And so that's how we are thinking about it, critical partnerships. We have laid out new partnerships around quantum computing domain, where we, by ourselves, cannot do everything. We are an application at that solver and algorithm development layer, but we need to work with partners, including middleware partners.

Some of them are Strangeworks, for example, ClassiQ, and a few others that will be coming up. And then hardware partners, where we are validating and verifying and benchmarking our solvers and our algorithms on different hardware modalities. So that's the final area of investment that we're doing.

Closing

WB: Yeah. Thanks for listening to this episode of the Aerospace and Defense Technology Podcast. You can subscribe to us on all podcasting applications, or find new episodes posted at techbriefs.com/podcast.

‍

Listen to the Podcast:

🟢 Spotify - https://lnkd.in/dkePcg6U

‍🟣 Apple - https://lnkd.in/ddYBcyQq