Podcast with Vishal Shete - Terra Quantum AG

My guest today is Vishal Shete, director of strategy and product at Terra Quantum AG. Vishal and I spoke about why they chose to build their own quantum cloud, the kind of quantum solutions that are moving from prototype to production, and much more.

Listen to additional episodes by selecting 'podcasts' on our Insights page

The full transcript is below

‍

Yuval: Hello, Vishal. And thanks for joining me today.

Vishal: Pleasure to be here, Yuval.

Yuval: Great to have you. So who are you and what do you do?

Vishal: Well, I'm currently a director of strategy at the quantum computing company or quantum technology company, Terra Quantum. I've been in the quantum world now for about three and a half years. It's been a fascinating journey.

Yuval: Tell me a little bit about Terra Quantum. Where are you based? How large are you? What do you do?

Vishal: Yeah, so Terra Quantum is a Swiss headquartered quantum technology firm. So we're about 80 people. We're actually growing closer to 90 people with offices in about six locations around the world. What we do can really be summarized into three things. We provide quantum algorithms and software as a service, we provide quantum computing systems as a service, and we provide quantum security as a service. And often, we combined our algorithms part and our quantum systems part to offer end-to-end solutions for customers at times.

Yuval: So, what does that mean? If I'm an enterprise and I want to get into quantum, and I have an optimization problem or a chemistry issue. I come to you, what can you do for me?

Vishal: Let’s assume you already know the quantum problem that you've got and want to solve better, either faster or get to a better answer. We'll work with you to really understand where are your limitations today. We can help develop an algorithm. Our algorithms team has some unique capabilities that develop very efficient use of quantum circuits, and we can develop these algorithms for you to solve your problem. And we can even execute it on our quantum hardware, which is a combination of virtual and simulated hardware, as well as integrated with other QPUs, quantum processing units within the system. So, in essence, what we would do is solve a problem for you with the best capabilities that are existing today in, I guess, the quantum and quantum-inspired space.

Yuval: Let me play back what I heard: I come to you with a problem. You guys have been doing this for a while. You've got 80 or 90 people. You've done optimization problems before. So you'll write quantum code for me that solves my problem after you understood the problem. And then you help me run it. Is that about right?

Vishal: Yeah, that's about right. And we'll also compare it against your best way of doing it. So what do you do in your industry, in your business today? And what's the difference that we'll be able to deliver in terms of performance benefit, but importantly, in terms of a business benefit for you? So what does a better optimization, what does a better global minimum mean for you in business terms?

Yuval: How does it work on an IP perspective? Do I get to keep the code that you write for me? Can I change it later? Do I have rights to it?

Vishal: There are different ways in which we work with various customers. So at the most basic level, we've got our IP in terms of the core of the Terra Quantum IP in these kinds of instances is in kind of the algorithm development part. So we've got some key components of optimization algorithms. One of them is what we call QNC, which is just quantum and coding. So that's our core IP. We customize that and apply that to your optimization problem, and we develop an end-to-end solution for you. And in addition to that, we've got our IP around the actual physical hardware and the execution environment that we run it on.

So in terms of the way we engage with customers, it's either a bespoke solution for them. So for customer X, we develop a bespoke solution where we hold the core components of the IP, but we do the end-to-end solution, and that's your solution. Or we look at developing solutions with them that we take to market together. So we develop a solution for an industry problem that they are aware of, that we're aware of, and together with them, we take it to the rest of the industry as well.

Yuval: When you mention quantum systems, does that mean that you have your own quantum cloud?

Vishal: Yeah, that's right. So we have a system called QMware, an entity called QMware, which is our quantum cloud essentially. And so this quantum cloud today is got a very unique positioning in the sense that it is in-memory computing with high-performance computing and virtual QPUs. So we've got 40 qubits, which we virtualize. We run that in a very tightly integrated way with really powerful best-in-class high-performance computing systems, which we can deliver end-to-end results for customers. But in addition to this, through this cloud, we'll also be able to integrate quantum processes as they mature. And we do integrate quantum's processes today, and we'll be able to execute algorithms and programs on this cloud, powered by the most powerful quantum systems when they're of a size available to create enough performance advantage. So in a sense that customers, they develop their program or algorithm once with us, and that's it. From there, as the hardware materials, we take care of it from that point onwards.

Yuval: I find that really interesting. How does a 100-person company conclude that it needs its own cloud as supposed to: “Hey, Mr. Customer, we can work on Azure, or we can work on AWS Braket, we can work on IBM.” Why do you need your own cloud?

Vishal: To be clear, we're not tied to our own cloud. That is one option for one option, one platform for executing algorithms that we developed. QMware was developed as a joint venture between Terra Quantum and another company called Novarion, which has got a pretty long and well-established history in Europe, especially around high-performance computing. And the reason we developed this cloud is that we saw a gap in the market for really well-integrated virtual QPUs or QPUs, quantum processing units, in general, with high-performance computing. So the in-memory compute layer that we have that integrates these two components together is really quite unique. And it's not something we can achieve or anyone one can achieve through other environments. So, that's what gives us unique advantage. But as I said, we work with the rest of the ecosystem a lot, and we're happy to continue doing that.

Yuval: May I ask what types of quantum processing units you have in your cloud?

Vishal: So we've got the virtual QPR, which is 40 qubits. Forty qubits which are, let's say, noiseless and because they're virtualized and fully interconnected. Which is immensely powerful. So 40 logical qubits, which is quite powerful. So, in addition to that, we work with the superconducting systems and some of the quantum annealing systems that exist, and we can execute on those as well.

Yuval: What are the main types of problems? We spoke a little bit about optimizations, but what other problems do customers come to you with?

Vishal: So optimization problems are a big part of it. So essentially, there are three areas that we see. So optimization, machine learning, and simulation. So optimization problems are quite unique and actually quite well suited to pretty much every industry. And I guess what we do differently in the optimization space is we're not limited by a QUBO standard, which a lot of people are, in which you have to be limited to a binary variable, to binary type problems, and you have to then discretize what is essentially a continuous problem and so on, but we're able to deal with a large set of variables, and we are able to deal with large type of variables in the optimization space.

Which gives us a lot of capability in addressing different types of optimization challenges. But in addition to that, we've got machine learning problems where through a hybrid neural network, we speed up the way in which we help train our models compared to classical approaches, but also improve the accuracy of predictions and simulation, where we speed up simulation system, simulation of various systems. So relevant for spaces around computational fluid dynamics or around even the Black–Scholes model and finance and things. So relevant for partial differential equations in different contexts.

Yuval: Very interesting. When you think about your customer base, how much of it is exploratory: “we're just getting into quantum, we'd like to compare the performance with the quantum system with the classical system” and how much is in production: “I have a quantum solution that I didn't have before, and now it's running, and it's part of my production operations?”

Vishal: And it's a really interesting question. Because we're very much set on this hybrid framework and doing the best that we can with virtualized qubits and high-performance computing and stuff, what we've noticed is there are some instances where we can create performance, which can't be achieved otherwise, that actually allows us to productionize solutions at large scale, especially with the algorithms that we've got. So there are I'd say about 20 to 30% of what we do is actually building that production, moving beyond the POC into a production built. Whereas the rest of it is working on a POC with a view of taking it into production. So I'd say there's a big drive to not just do toy problems and not just do a problem that we can then do a press release about or whatever and say, "Okay, organization XYZ is now in quantum because we did a POC." We actually want to do something real and tangible with as near a term impact as we can create with whatever structures that we have.

Yuval: Are you able to share names of customers that you've built quantum solutions for that are in production?

Vishal: I'm not able to share names of customers, but I can give you some color on the types of customers that we work with. So in the investment banking space, we have solutions around a particular problem called collateral optimization. So this is one of my favorite moments in quantum came in the collateral optimization space because it's really relevant for any trading function where anyone that has a margin requirement that needs to be met needs to manage the amount of collateral and what collateral they need to post to meet that margin requirement. So what we do is optimize the way in which you meet your margin requirements in the cheapest way possible for an investment bank or a trading function that may not be in a bank. Why it's really exciting is because, with the solution, we move away from just talking to the quantum team and quantum experts, which we do, of course.

But in addition to those, we're talking to and implementing solutions with heads of trading, heads of liquidity, heads of capital at really global investment banks. One of them is about 400 billion in terms of collateral that they manage. And so that's really quite exciting because those people don't really care whether you're executing it on your calculator or whether you're using a quantum system, they care about the output that you get for them. So they're quite an interesting bunch to work with and very impatient. As you know, most trading people in investment banks tend to be.

Yuval: Is it a time-sensitive application? Do they need responses within seconds or less than that?

Vishal: No. So, this is not one of those where speeding it up much faster will create a better value because it's done typically once a day when you look at everything that you have to post to and try and find an optimal solution for it. So it's done once a day, and it takes about an hour today in terms of processing time to do it. So the value really is in finding a better answer rather than finding the same answer faster, which is what we are able to demonstrate, which is highly exciting.

Yuval: Roughly how long has it been running in production?

Vishal: To begin, where we're implementing it in production now. So it should be in production in the coming months. We've been working with customers. The first POC was done at some point last year. And that's been the journey since sometime.

Yuval: You mentioned that you guys are 80 people growing to 90. I would assume that most of them are engineers. And so you have many man-years of experience working on quantum systems. What would you like to see hardware-wise or software-wise to make your jobs easier?

Vishal: It's a very interesting question. I think for us, what would be one of the key things that we need to solve from a hardware perspective is if we can get to a better quality of qubit. So once we can do that, I think that is hugely powerful. I think more than the number, the quality and the interconnectedness of qubits within different systems is super important for us. So, for example, we're working with 40 logical qubits, which are fully interconnected. Now getting to that stage in a real system is going to be great. And getting beyond that will be hugely powerful. I know others have got more than that in terms of physical qubits, but when you look at logical qubits, it's much smaller set to that. So error correction and managing the noise within qubits is a hugely important task in the whole space, I think.

Yuval: You probably speak with a lot of corporate customers about quantum. Do you think the industry is overselling the capabilities of quantum today, or maybe it's underselling, and actually you can do all of these wonderful things right now, and corporations are just not aware of it?

Vishal: I think there's a huge potential in this space, as many people who are listening to this are probably well aware of. Now, I think the problem in terms of how it's communicated to the corporate customers is that it's not quite communicated in the right language. So making it more operational, more tangible, more concrete in terms of having a clear use case in application view is quite important. And I think we're moving towards that a little bit, but I think there's still some way to go where, in my view, you get to a world where when you're talking to customers you're talking less about qubits and different types of algorithms that you're using, but more in terms of outcomes that you're achieving for them. And then that's where I think the industry should evolve where companies are now focused on solving particular applications and use cases and creating value in this particular space, as opposed to being a quantum company, which is such a broad thing.

Yuval: If I try to paraphrase what I heard from you, and maybe I do it clumsily, you would go to a customer and say, "We harness the power of quantum computing to give you a better collateral analysis solution."

Vishal: Exactly.

Yuval: And in your view, what would help accelerate the adoption of quantum tech with corporations?

Vishal: So it's pretty much that. It's pretty much being able to be concrete about this and being able to translate the value that can be derived through this in a way that people can understand. But also, I think on a more sort of technical note, I think it's important for us to be able to continue working with the best in class HPC solutions and that exists here in HPC. When I say HPC, it's broad bracket, including FPGA's and GPU's and everything else. Figuring out how we can work in the best way in integrated environments with all these things together is super important. So I think those two things I'd say, integrating with the best of what we've got today and communicating the value in a way that can be understood easily.

Yuval: As we get close to the end of our conversations, I just want to revisit a point that we discussed earlier and about your private quantum cloud. Do you expect to continue to do that, or do you estimate that in two or three years, there's just going to be clouds from the major vendors that are going to be good enough, and you'll abandon your private cloud effort?

Vishal: No, we definitely expect to continue to do that. We expect to continue to build our cloud, and we potentially will engage, we'll integrate with some of the other vendors as well. So have our cloud accessible through other clouds that exist and vice versa. So, that I think it could be a path that could be quite interesting. By the way, I know we're coming up to the wrap-up, but one point I'd like to just touch on quite quickly is our work in the quantum key distribution space, which has actually got quite a bit of coverage in some of the major media outlets, which is tremendously exciting because actually, firstly I think quantum along with the optimization space, my views that quantum key distribution and quantum security is one of the earliest real-world and large-scale applications of quantum technologies that exist.

And what our scientists have been able to do is develop a quantum key distribution protocol, which is not dependent on new quantum infrastructure in the telco system. So, which means we can have a long-distance secure quantum key distribution up to 40,000 kilometers, which is huge because that allows you to set up a full network without having to re-engineer existing infrastructure. So, that's another big project that we're working on with several customers on.

Yuval: Thank you very much for bringing that up. So, when you look at quantum algorithms, quantum software, quantum systems, and quantum security as a service, what's the best way for people to get in touch with you to learn more about your work?

Vishal: So the easiest way is either get in touch with me in LinkedIn, or drop me a line at vs@terraquantum.swiss. Happy to discuss with any interested parties.

Yuval: Excellent. Thank you very much, Vishal, for coming on the program today.

Vishal: Thank you, Yuval, been a pleasure.