SPEAKER_03 0:40

Hey Tiffyin, how are you doing?

SPEAKER_01 0:42

Hey Patrick. I'm doing good. Looking forward to another great episode of Entangle Trans.

SPEAKER_03 0:46

I don't think you'll be disappointed because we're filling in a missing spot. So we're joined by Abby. Abby, do you mind introducing yourself to our audience?

SPEAKER_00 0:54

Hey everyone. I'm the founder and CEO of BQP. By training, I'm an aerospace engineer and a computational scientist. So not a quantum physicist, but we'll talk about some very interesting things during the episode. Just my journey on how we got started. And these codes were written in Fortran 77 and C, taking six months on 250,000 computers for doing even one simulation. So I was like, how is this even useful for anybody trying to make decisions about design or things? And I wanted to solve that problem. So my first attempt to solve that was with Okresh National Lab and Nvidia to bring these from CPUs, old legacy systems to GPUs. And what I found out that even on the largest supercomputer of the world, we were taking, you know, only utilizing 15% of the GPU power. So about 85% getting wasted because of the memory bandwidth that GPUs have when you're trying to put very large complex systems into like GPUs. So that gives the insight that it cannot be just ported, it has to be refactored from scratch. And that's where we started BQP. And I'll talk more about what BQP is doing in the following uh discussions. But that's that's the background that led to the pieces of the Peter.

SPEAKER_03 2:25

Well, I mean, uh it's it's it's a cr it's an interesting space because there's lots of people making impacts that aren't quantum physicists. But it it was that way for a little while. There was a rarefied space in time, but I think we're moving past that. And I think that's that's required because there's just not enough quantum physicists to do what we need to do.

SPEAKER_00 2:45

Yeah. Um I also believe that physicists think differently than engineers do. So we need both. Um, the word need both.

SPEAKER_03 2:55

So uh I mean B BQP is solving what problem in the quantum space for those that aren't familiar with with what you've just said.

SPEAKER_00 3:04

Um yeah, so we are on the application and the algorithm part of the stack. Um if I have to describe um in in very simple words what we do is that we are putting uh physics into the decision and control loops that was missing because simulations take a lot of time, as I just described. Uh, and that's important for many, many industries, uh, starting from where we have started, which is aerospace defense, space, uh, now going into energy systems, uh, semiconductor, uh, and then beyond engineering, um, where physics is equivalent to math, and then math is what goes behind finance and biotech and others as well. So that's that's the um problem we are solving for.

SPEAKER_03 3:49

Cool. Um is there something that you wanna I mean talk about as far as you guys offering a product already? You're you you're doing consulting mostly. What is your space? Uh it sounds more like a consulting play.

SPEAKER_00 4:07

Um we actually have a product, uh, one core engine, uh, which utilizes uh two forms of quantum algorithms. One is quantum inspired, which are basically under the solver called Quantum Now, and they run on existing CPUs and GPUs. And basically uh helping to understand which problems are good to be solved with quantum and which are not, which both both the learnings are very critical. Um, and at the same time providing value to end users today, uh, not only by making them quantum ready, but also by solving actual problems they care about to be solved today. And that actually bridges the gap of the big question that always I've heard from our partners in quantum hardware, middleware, where they go to customer and they are like they ask, okay, what's your most complex problem to solve? You know, and they're like, uh, but what quantum is good at. So there's that tension that always lies in and you're not able to come to intersection. So we are solving that with first principles. And that allows us to then map it to our next solver, which is called Quantum Max, which is basically with quantum native algorithms running on actual hardware. And we're currently performing RD and working with partners in the quantum hardware and middleware space to map those and understand resource estimation and when they will be ready for prime time.

SPEAKER_03 5:33

I think that's that's definitely an important space.

SPEAKER_01 5:36

And and this this is a very um old quote unquote problem, but at the same time, I think it's beginning to uh uh emerge as a critical problem, which is having the proper way to determine like which problem is suitable for quantum and which not. And I would even go into more detail like what part of a problem is suitable for quantum and which could still be solved, right? And then I think the other one that's also um pretty difficult to answer is how do I represent my problem so that it could A be solvable with quantum and B take the most out of what any quantum hardware would would be, right? So I I think that the work that you folks are doing is is is becoming increasingly important because even if we would have like today uh stable, fault tolerant quantum hardware, like these two questions would still be very difficult to answer.

SPEAKER_03 6:47

Yeah, we've we've had some clients who, not clients, we've had some guests. Guests who have just who have basically proposed that their problems are quantum inspired and best solved that way. And they they don't really see quantum hardware as required for it. And so sometimes making that decision is the right decision. And and the reason I say that is because we've had all this um back and forth of, hey, we have quantum advantage, and like then the supercomputers say, No, you don't. We we can do that just as well. And so we you're kind of removing some of that noise off the table.

SPEAKER_00 7:24

Yes, and that's where I'm big advocate of not thinking of it's not a world of quantum inspired versus quantum native. They're part of the same continuum and it's an on-ramp. One cannot exist without the other, even in future. Even with stable qubit, if you have to do a production level fluid dynamics, you you're not going to directly put on a quantum hardware. You're going to do some checks and balances. Is my mesh right, is my inputs and outputs right and everything. And you will do that on a quantum-inspired form, right? Um, and then when you are ready, then you will run your production level uh simulations or whatever, your calculations on a quantum uh hardware. So so that's where I go back and I am telling and talking to people more about this and through your guys' platform would love to also put this thought process that it's not a battle between quantum inspired and quantum native. They're part of the same continuum.

SPEAKER_01 8:23

I mean, I I I like that you I like that you mentioned fluid dynamics because I think this is also one area where we will see synergies between quantum inspired and quantum native on one hand and AI on the other hand, right? I'm just thinking about the recent results that were announced regarding the Navier-Stokes equations, right? And the uh singularity that was demonstrated uh using AI models. I I think uh we've been talking on the show a lot about, yeah, how will quantum help AI and how will AI help quantum. I think one area where this will happen is literally identifying the slices, the pieces of a bigger problem that could benefit from either quantum-inspired or or or or quantum. Uh you mentioned something in the beginning, though, that I would really like to get a little bit more detail, right? As uh so when you propose or when you run solutions on on classical computing, you also have like uh a layer that would improve the capability of using the existing hardware. Is that part of your um uh approach, or you're relying on on some kind of third-party libraries to do that?

SPEAKER_00 9:46

So that's the core engine we have developed, our product BQFY, which is which has multiple layers. So it has the understanding of a mathematical problem and how it efficiently maps to the underlying hardware, whether it is a CPU, a GPU, an FPG, or a quantum computer. Now, how we do that is basically the biggest challenge which we have seen with classical algorithms and traditional approaches on a GPU is that its memory GPUs are memory bound. So if you have a very large problem, what GPU has to do, it has to keep communicating with a CPU uh to support it. And that communication basically reduces all the benefits that a GPU would have. And that's where I mentioned in my anecdote of the work I was doing uh with Oak Ridge and Nvidia. Like we found that. And now, more so we have proven that that that's the case. So, what the platform is doing is that it has, for example, an optimization capability. So it maps any kind of non-convex optimization problem much more efficiently than a genetic algorithm would. Uh, it's it's inspired by genetic algorithm, but it's a more GPU compatible, GPU efficient part of it. So we are utilizing the maximum course required. And that was made possible through the quantum inspired or the tensor network mathematics that we brought in. Similarly, we are also looking at a very interesting problem is uncertainty quantification from engineering industry. I mean, it's not just in engineering, it's they are in many other industries. Finance people know it more as Monte Carlo, but it's one and the same thing that we are doing. Now, that problem has the classical part to it, meaning it has to build all the samples classically. So that's the part we have to do on CPU, GPU. The Monte Carlo part is what we do on a quantum computer. So that's the heterogeneous thought process that we build out only after working with many aerospace defense companies, saying that, oh, beyond simulation, this is what we want. You know, and and the best part is Monte Carlo type of algorithm maps to quantum very well. So, which modality does it map to well is needed because this kind of algorithm, at least the version we have implemented, has high entanglement. So if you're putting on a system which does not have, let's say, all-to-all connectivity, then what happens is it needs to have a lot more swap kits. And that reduces the benefit that you will be getting because the more the circuit depth, the more the error you are introducing in the system, and it will take you know much more logical qubits to do it. Whereas if you're mapping onto something which is, let's say does not require that many swap kits through its architecture, uh, we have showcased that that's more efficient in terms of accuracy that it will get. So this kind of learning is not something that I can come in and say to my end users or my partners, oh, we are hardware agnostic. There's a reason why we are hardware agnostic because we are putting the pieces of the classical, the compute, and together, and then mapping each problem by its merit. So we are problem-driven rather than solution-driven, and that's the mindset I think engineering gives to us, which I was mentioning earlier.

SPEAKER_03 13:07

Yeah, this must be an exciting time for you because you mentioned, you said something that I really like and respect is you you qualified the type of qubit you were talking about. You said logical qubits. And we we've just emerging from a period where people would say qubit and would mask whether they were talking about physical or logical. A lot of the hardware vendors would say, oh, we have a thousand qubits, but you know, that never so we're we seem to be coming to the point where when people say qubits, they mean logical, which it seems like that would be very important for the work you're doing. Because you don't care about physical qubits, you care about what qubits can I use that actually do the work.

SPEAKER_00 13:46

Absolutely, because and and the the challenge that we are currently solving with some of our hardware and middleware partners is what's the like if they say that they have X number of physical qubit, and if I have to get to a logical qubit thought process, then how many of those qubits will in the logical qubit actually be utilized for compute, and how much would be required for the rest of the error correction or mitigation? So that understanding of both, and but more importantly, the mapping between a physical qubit and logical qubit is very critical. Um, and it's not easy, it's something that we have to dig in and really understand, and and and it it is only going to come by trial and error, which is why I go back in saying that the quantum inspired is helping us to understand which problems people will pay for in future, but then all these work is also getting influenced because of which problem to solve. Not what's my best algorithm and what's my best hardware, and everybody should use it. That's it should be problem-driven.

SPEAKER_03 14:48

We're still in that fuzzy area where I I I assume, and and assumptions are deadly, that in 10 years there there will be clear things that are just better on quantum hardware. But we're it's still fuzzy, as I said, because of that back and forth of we have quantum advantage, no, you don't, you know. Um so so I think there's there's this time period we're in now where quantum inspired can arguably be just as good as quantum because we have such the NISC kind of computer, noisy uh quantum computers. But I'm hoping we emerge from that by the end of the decade. And with when Shore's in our back in our rearview mirror, which is kind of almost hard to imagine because it's been overshadowing and forcing the money into this system for so long, um then the the ability for quantum to do things that no other system can do will kind of shine at that point.

SPEAKER_00 15:45

Two points I want to make on that, Patrick. So number one, um I I will say that um I wouldn't categorize that we are getting advantage from quantum inspired. I would say that it's we have you're spending trillions of dollars on HPs and GPUs today. So quantum inspired is helping to make that investment worthwhile.

SPEAKER_03 16:08

Exactly.

SPEAKER_00 16:09

And it's a bridge. Yeah, it's a bridge. It's a bridge.

SPEAKER_03 16:12

Yeah.

SPEAKER_00 16:13

While the NISC is helping us to I I even say that like quantum algorithms running on a simulator is helping us to build that mapping, which which Scipion mentioned earlier, right? Like that mapping is very critical. How does an algorithm go to a quantum uh hardware? How will the circuits work? How will the noise models work? And all this thing together, what's the resource estimation? So those are the things that the niche should be looked at, not from a quantum advantage point of view. And then comes the third part of it, which is where the advantage will show up, and that has to be logical qubit error, but we cannot wait for it because the wrong, the the way this field, and since I've entered this field six years ago, people have been constantly just talked about, as you said, oh, this is at quantum advantage, or this is not. Yeah, everybody's claimed. Nobody's talking about from a viewpoint that I'm showcasing quantum advantage for this specific problem on this specific system with this specific, you know, configuration. And that's the right way to go about it because you will see some quantum advantage emerging in the next two to three years. But that doesn't mean that quantum is ready for everything. Because it's a roadmap, it's it's a journey that we have to cross.

SPEAKER_03 17:30

When you have a hammer, everything looks like a nail. Exactly.

SPEAKER_00 17:33

So that that that I think is very critical to understand. Um but the the second point um here, you know, on on your response was we we have to be also thinking at the same thought process in terms of that NISC has been and and has been driven by the fact that it should show quantum advantage. AI didn't show advantage for decades, right? Before the chat GPT, people think AI is so new, but it's not. We all know that the theory came from in the 70s, 80s, maybe before that. But this is the point in time where actually I'm able to see the quantum for AI and AI for quantum story written on the wall. And and the recent, if you if you follow DOE and the Genesis uh projects that are coming out and the calls that are coming out, is asking how AI will help in developing those quantum algorithms which it will be useful for. And that kind of AI agentic thought process, I think is going to be very powerful in helping companies like us who are developing those algorithms and need to do this manually. You know, I've been doing we've been doing this manually for five years, like every test. And if but if so, if there's things that could help move us forward, we would be sitting with not just Short's algorithm, but many mature algorithms that we'll be talking about.

SPEAKER_03 19:05

You move up the food chain basically and you know, get get out of the grunt work. Uh, what is it? Dangerous, repetitive, and boring are the things that that robotics and AI are are meant to solve. So very cool. Um do you guys uh go to go to conferences? Do you you know, is there any place people might see you guys, or where where would where would we look for you?

SPEAKER_00 19:28

So we go to conferences, but we go to very less quantum conferences. Um we're trying to find a balance. We go to more so where our end users have been. So we have been prioritizing aerospace conferences like AIAA, ASME conferences, um, IEEE, but more from the applied side of things. Uh the only quantum conferences I uh have been, my team has been uh is quantum world congress, very regular. Uh so we will be there in about two weeks from now. Um we have very exciting uh things coming out uh during that time. Cool. Um and then we are also looking uh forward to Q2B um in in Chicago later this year. Um so we are trying to find the balance because there are too many quantum conferences, you get to meet the same people.

SPEAKER_03 20:17

No, no, I I I I think I think it's actually better for everyone if if companies like you are represented at at the uh the non-quantum conferences, so people because because then that's how I got into it, um, as I was uh discussing with you before we started showing. Cyprian uh I I I saw a conference where one qubit out of Vancouver talked. It was probably 12 years ago, uh but and they were doing quantum in inspired. And then I started talking to Cyprian, and then we just every time we saw each other, we were talking about this stuff. So it's important to get that message out there. So uh are you finding that it's in the consciousness of people in the aerospace industry that quantum is a force and something that's coming to disrupt? Or are they like, well, what's quantum got to do with what we're doing?

SPEAKER_00 21:03

I think we in the past three years we have moved from that what quantum has to do with this to even conferences like ARWA, which is I would say the premier aerospace conference in the world, have a section talking about the future of compute, and there's at least two to three papers on quantum out there. So that's that's a big progress I've seen in the past two, three years. And and I think people are just generally aware. Now, aerospace defense is uh different. Just talking about aerospace, um, the world of fluid dynamics dictates this field, and the fluid dynamics, the CFD community. Uh, I would say 2014 they wrote about quantum in their roadmap. So there is a NASA CFD 2030 vision document that you can Google and see it. Even in their first or in their second version of the document, they actually wrote that quantum will happen and fluid dynamics and uncertain quantification will quantum. Will happen, but they just wrote it. They they didn't expect companies like BQP to come and you know actually make it happen and be in those conferences and presenting in those um CFD 2030 um like sessions and talking about quantum, but I think I've seen in past two, three years change. In generally, I would still say that the end users, leaving the top few aside, you know, you you always have an industry where they are already thinking about quantum, they have quantum teams and everything. Anything from those top, even like half percent of companies, they are working with us purely because they require physics in their decision loop. So faster simulations, more accurate simulations. Um how we are getting there is not really relevant to them. Um, and that's something that we have tried to insert quantum inspired, quantum inspired quite a lot. But I think we have uh being very honest and transparent, um, I think that has played kind of against us, you know, just mentioning quantum inspired, because then we go into this deep journey with engineers. What is quantum inspired? Yeah, how are they different from classical and and different things, right? So we can make our lives very simple, and we have tried to make that that people care about solving their problem. How is not the most important part?

SPEAKER_03 23:22

Yeah, educating the space is always difficult.

SPEAKER_00 23:25

Exactly.

unknown 23:26

Yeah.

SPEAKER_00 23:27

So you you show the benefit and then you tell them, oh, this is how we got it. Then it's proven. Engineers love proof. That's true.

SPEAKER_03 23:36

Uh in the early days, Cyprian and I were like, well, where are all the algorithms? Because we got Shores and Grovers, but uh it seems like that we've moved beyond that like big name name algorithms. Um, you guys are probably not dependent on that, or or you see the the future differently than we're gonna have the this this list of well-known algorithms that we deal with. Um is that is that was that just the wrong way to think about it, you know, five, five, six years ago?

SPEAKER_00 24:05

Um again, I I think I'll go a step back from there who was talking about it and how those like the early companies were like, as I mentioned, run by quantum physicists and quantum engineers. That's a different thought process. And I'm never going to say that's not the right thought process, but I think But it's not the way the industry is moved. Exactly. If you're talking about the for me, the most important qualification is how is the end user thinking about it? And end users think about things in a system system like engineering point of view. I call it system engineering, but it's like algorithm is only a piece of the puzzle. What comes before that, what comes after that, how will somebody utilize it? For us, the biggest learning that we have changed and which has changed the trajectory of BQP is our integration into so just as a background, we were working with several tier one OEMs in aerospace, come aerospace. They used our software, quantum inspired, they even gone to the quantum native side of it, saw what how quantum will work, so all of that was great. But then when it came to okay, we have done this POC, we have done this pilot now, it's exciting, let's go to the production level. And we there we realized that the way we have built our software layer, the platform layer, is not something that engineers can use readily. So that learning, that feedback came in. What we did after that was we made it seamlessly integrable into something like a MATLAB. So now we are natively integrated into MATLAB, which is where engineers live.

SPEAKER_03 25:40

Yeah.

unknown 25:41

Right?

SPEAKER_00 25:41

So engineers will be using the same way as how they were doing things, no change in input, no change in output, no change in infrastructure. But they can now use the power of quantum, which is what they what matters to them.

SPEAKER_03 25:53

You're meeting them where they are.

SPEAKER_00 25:55

Exactly. And that has changed and transformed. So post that we've done integrations with Python, with Julia and others, and that's like I said, is the system way of thinking. You just don't cannot think of quantum algorithms, and for many companies, I do advise, especially young companies that are coming up, quantum algorithms and algorithms cannot be your secret source. Somebody, forget somebody, even AI will find it, you know, better algorithms for you. So you have to put that as part of your one layer and then build other pieces around it. Cool.

SPEAKER_01 26:30

Now, one area that I would like to also uh uh understand a bit more um regarding what you folks do is you mentioned about estimation. Um so can you tell us a little bit about um what are you doing in that in that space, right? Especially now that we're starting to see, and we'll probably see a significant increase in the next two to five years of uh quantum hardware that can actually run um uh error corrected logical qubits, right? Um, estimators and estimation, I think it's another very important aspect because it's not only the algorithm, it's not only the problem, right? But it's going to be all right, can we grab the right number or do we have enough qubits to do this, right? So estimating is is is is critical. And I think I heard you mention that that you're doing work in that space as well.

SPEAKER_00 27:30

Yeah. Um three to three years ago, I want to say we for fluid dynamics, we started that, where we knew that if we have to come anywhere closer to providing quantum advantage for fluid dynamic type of simulation, especially for large aerospace systems, oil and gas energy, it's a long roadmap. And people are not looking from us. And if you have to be practical, we we cannot be showcasing um quantum advantage for a really long time, at least three years back, I would say that was true. And now the timelines are closer, but still like a few years away.

SPEAKER_04 28:09

Yeah.

SPEAKER_00 28:09

But what can be additive to the literature and to the industry thought process is what type of quantum hardware with what resources. And when I say resources, double-clicking on that means not only a qubit count, but also the decoherence time, which is decided by you know how much uh circuit depth you have in your algorithm and two qubit gates you have in your system, um, and and how much error correction or error mitigation do you need. So categorizing all of that became our thought process. So if you look at some of our publications that we have brought out, even in AIAA, in other conferences, has largely been focused on moving in that roadmap forward. And that roadmap will change. What I published three years ago versus what we will be publishing this year or next year will look different because the systems are changing, our algorithms are improving. And one of the biggest bottleneck that I've seen goes back to Sitran, what you were mentioning about you know how will quantum and HPC work together. Fluid dynamics is a great example for that because it's highly non-linear partial differential equation. So part of that you have to do is you have to bring it to linear if you have to put it to quantum. You bring it to linear, that all of that is a classical process. Once you're ready to put that into quantum, then from there on you basically have to make it a Hamiltonian system. So Hamiltonian can be solved in the NISC era through like deviational circuit and in quantum native era through like HSL type of algorithms. But that's how things work. Now, when we started doing that, because the approach is how we are linearizing the problem, is still your classical method of finite element, finite volume type of methods. What ends up happening is your circuit depth is very large, meaning I can only go to certain number of qubits, not because I'm limited on qubits, because I'm limited by the circuit depth of what I can run. In and I'm when I'm doing that on a simulator, I'm doing that bounded by the memory. When I'm doing that on a quantum hardware, I'm bounded by decoherence time. So what we have been then learning from this and publishing about this, what we have been focusing on, how can we reduce circuit depth? So then we worked with Classic and Nvidia to be able to showcase the reduction in circuit depth so that we can go furthermore and then furthermore. So this kind of resource estimation, if you look at some of our recent work in uncertainty quantification, that's really where we are investing our time and effort. So not like we know that the chase is long, the roadmap is long, but these are the milestones that are very key, not just for our development, but to tell the industry, to tell the quantum industry, our partners, these are some of the things that we should be focusing on more, and not just qubit numbers. Because I think largely we have just talked about how big the systems are. It does not matter. It's the circuit depth that is restricting us.

SPEAKER_03 31:17

I mean, that's similar to the classical world. We, you know, there was a time when memory was the bottleneck and then storage was the bottleneck, and bandwidth is the bottom is always another bottleneck. Uh, and and we forget, you know, how hard it was when we were fighting with the old bottleneck, and we just think about the new one. That's that's how progress goes. So we we've been at this for a little while. We still got a little bit of time left. Is there anything else that you wanted to uh you know share with the audience and us that um we haven't discussed yet? Because uh we want to make sure that your message gets out.

SPEAKER_00 31:49

Yeah, I I'd love to share a few examples of how industries are thinking and and progressing from this continuum I mentioned from quantum inspired to quantum native. So uh the the first example I'd love to talk about is a drone company called Modovolo. So they again found our software uh from MATLAB. We are in the same vicinity here in Syracuse. Um they are they're a really awesome, highly technical team. Um one of the problems they were trying to solve for for for a while is uh their propeller design. Uh they sell to the defense industry, so they're restricted by who they can source their propellers from. And this is a purely engineering design problem. So they found about our software because they were getting restricted by the classical genetic algorithm. What ended up happening is they used in their engineering flow half a line of code changed, they used our software, and they went from not just doing simulation, but actually uh getting to a patent pending 3D propeller that they actually 3D printed it. Cool. Uh tested everything. And that journey in itself has made them quantum ready. Now these guys are thinking with us is okay, this is exciting. How we can go from one system like a component like a propeller to an entire system design. We are pursuing different uh things there. But also the question then immediately comes up is what would the benefits would be like if we were doing this with a quantum hardware? So as an optimization problem, as an engineering design problem. So this is very exciting work, you know, and and a way for industry and end users to get started in the journey with quantum. More lower lift, get the performance improvement, and then you will be more excited and motivated to look into quantum more seriously.

SPEAKER_03 33:51

Very cool.

SPEAKER_00 33:52

The other example I want to uh talk about is the space work. So that's uh SBIR that we have recently won. The work started before that in a program called SDA Tap Lab uh in Colorado Springs, where we worked with space companies. And there the problem was your again, your classical solvers were uh coming from the viewpoint that they are either highly accurate or they are fast but inaccurate. So some balance has to be striked there. So there we developed our own physics AI model, physics informed machine learning to be more precise. But when we trained the model, the model was very expensive. And the objective that was set by Space Force was they want to put this at a tactical tactical edge. So that's where quantum came in, uh, more quantum-assisted machine learning, where we replaced some of the uh parts of the neural network with the quantum layers. And what it did was it reduced the model size from close to around 6.9 uh gigabytes to three megabytes.

SPEAKER_03 34:58

Wow.

SPEAKER_00 34:59

And that actually was able to get deployed on a Nvidia Jets and Nano. So training is quantum plus classical, inference pure classical. And that's how we basically got awarded the Space Force SVIR to now put things in space at a tactical edge.

SPEAKER_03 35:21

Nice.

SPEAKER_00 35:22

With these two examples, what I'm trying to get to is that quantum, and I I like borrow this uh quote or thought process from what IBM have mentioned, don't think of quantum as a car versus a fast car analogy. Think of it as a car versus a boat or an airplane analogy. That you are able to now get to certain places that was not possible before.

SPEAKER_03 35:49

That's good.

SPEAKER_00 35:50

So that's where the learning has to be, that it's not about making just simulations faster. It's that you're removing the memory bottleneck, so things are getting you're utilizing the full GPU space, hence things are getting faster. In the case that the second one in machine learning, because I reduce the model size, make it less memory bomb, that's why you're getting those kind of 2000 speed ups that we have, uh 2000x speed ups that we have seen. So quantum is an enabler to those performance, not the reason why you're getting uh that that speed up. So those are like the key things, and and our recent investment from IBM through their venture armp uh in BQP was was motivated largely. Uh they've written blogs about it, that somebody has had to think about from a problem point of view which problems are good and how to best utilize quantum in which part of the workflow. And and they like that, and largely that that drove their investment thesis in in BQP.

SPEAKER_03 36:51

Very cool. Well, I think we're gonna put a pin in it here. I would love to keep talking to you, but uh but we have to have to have some kind of line. Uh hopefully you'll join us again in the future, and uh, and what you're doing is uh is definitely tracking nicely with the way things are evolving in quantum.

SPEAKER_00 37:09

Thank you, Patrick, and thank you, Supin, for having me here.

SPEAKER_01 37:12

It's been a real pleasure.

SPEAKER_03 37:13

Thanks for joining us for talking about.

SPEAKER_01 37:14

Thanks, everyone. Bye. Bye.

SPEAKER_03 37:16

Thanks, everybody.

SPEAKER_02 37:17

Bye. Cybercrime is one of the biggest threats to businesses of all sizes and industries. With almost half a million open cyber positions, the problem is compounded by the lack of available talent in the marketplace. At AllStar Security, our elite team of highly credentialed experts collaborate with you to assess your current defenses and develop solutions tailored to your specific needs. With services ranging from cybersecurity education to advanced penetration testing and red teaming, you can start reducing your risks today. Visit Allstarsecurity.com and let's secure your digital future together.