← All episodes
Roman Orus

Episode · Feb 27, 2023

Roman Orus, Multiverse Computing

Why vertical expertise beats quantum-first thinking

Summary

Roman Orus, founder and Chief Scientific Officer, Multiverse Computing is interviewed by Yuval Boger. Roman and Yuval spoke about developing quantum applications for customers, whether the right solution for the customer comes from a vertical-specific company or a quantum-centric company, their operating philosophy, and much more.

Key takeaways

  • Multiverse Computing argues that solving a customer's actual business problem usually requires deep knowledge of that industry first, with quantum as one possible tool rather than the starting point
  • Companies that start from quantum hardware and search for problems to apply it to tend to produce weaker solutions than companies that start from the customer's problem and bring in quantum only where it helps
  • The bar for a quantum-based solution is not whether quantum works, it is whether the result beats the best available classical approach for that specific customer
  • Building trust with enterprise customers depends on being judged by business outcomes and domain fit, not by how advanced the underlying quantum technique is

Transcript

Yuval Boger: Hello, Roman, and thank you for joining me today.

Roman Orús: Hi. How are you?

Yuval: I'm good. So, who are you and what do you do?

Roman: Okay. Well, let me introduce myself. I'm Roman Orús. I'm a physicist. I'm the co-founder and chief scientific officer of Multiverse Computing, and I'm also a professor of theoretical physics at DIPC, which is a research center here in Spain.

Yuval: And what does Multiverse Computing do?

Roman: Well, let me explain a little bit about our company. At Multiverse Computing, we do software for quantum computers, okay? And in particular, we are focusing on applications for industry. We want to build applications for real problems, for useful problems, for different industry verticals.

And we want to solve these problems using quantum technology, okay? And also quantum-inspired technology as it is today.

Yuval: I saw a recent presentation that you gave, and you spoke about democratizing quantum.

Roman: Yes.

Yuval: Isn't it too early to democratize quantum? I mean, does—

Roman: No, it's—

Yuval: I think you gave the example of your grandma. Does your grandma really want to use quantum computers?

Roman: Why not? I mean, if she's given the chance, why not?

Yeah, absolutely. Well, first, to answer the question, I don't think it's too early. One of the things that you learn when working in such a disruptive field as quantum computing, and this is something that I've learned over the years, is that the future always comes faster than you expect.

You always hit it sooner. Many people think, "Well, it's still many years until we can do something useful, until this is mainstream." Well, you need to start thinking about this early, okay? Because otherwise you may be too late. So I don't think it's too early to democratize quantum computing.

At Multiverse, we are obsessed with different things. One of the things that we are obsessed with is solving real problems. Not academic problems, okay? Real-life problems, really the dirty stuff, and trying to reverse engineer and see how we could solve them with quantum technology.

But the other thing, which is extremely important, is to make the technology accessible to everybody, and this is why we talk about democratizing quantum. Quantum technology cannot be something just for quantum engineers, let's say at a bank, or for quantum specialists or quantum physicists.

At the end of the day, the people working at companies and startups like us have to make quantum technology accessible to everybody. We need to make it accessible to everybody. It's the same as saying, imagine that you want to switch on the radio and listen to some music, and that's it.

To switch on the radio and listen to some music, you don't need to be an electronic engineer and know all the details of how the circuits work and so on. This is exactly the same. We need to make the technology accessible for everybody, or otherwise this is not going to work.

So, yeah, I don't think it's too early. The technology is still under development, but it's going to be under development forever, okay? And we can already start to see the first useful applications in the history of quantum computing, and therefore it's just the right time to start thinking about this.

Yuval: The reason I ask about too early and democratizing, let's take an example. I think one of the examples that's brought up often with quantum computing, and I think you guys talk about it as well, is portfolio optimization, right? Take a basket of assets and optimize the allocation between them to achieve target returns and risks.

Who is your customer for such a program? Is it the 55-year-old employee who's starting to think about retirement in 10 or 15 years? Or is it the company that's managing the assets for that person?

Roman: Well, at this point it's the company. But still, if you go inside the company, it makes a lot of difference whether you are targeting the technical quantum engineers inside, or you are targeting the traders and the brokers, since you were asking about this.

Even the highly skilled quantitative analysts inside a bank don't need to know how to program a quantum computer. Not at all. You need to provide the solution in the language and with the tools that they already use in their everyday life.

Of course, they know a lot about financial theory, but they have no clue about quantum, and you have to make life easy for them as well. So in the particular case of portfolio optimization, what we did was develop a plugin for a spreadsheet, which is one of the typical tools that many people use at financial institutions, where you can download the data directly from Bloomberg and so on and just make it work as an optimization engine, where you don't need to know exactly what the inner workings are. Now, if somebody wants more detail, we can also provide that.

We can provide the explicit libraries, we can provide all the details of how the program is built, or the details about the code. We can also do that. Now, looking forward, right now this is commercialized by banks.

When you want to buy an investment fund or something like this, typically it's the bank that is managing its performance. But eventually, people also want to take care of their own finances, and in this sense, I can imagine that the tool we are building could also be useful for them.

Yuval: When you look at portfolio optimization, obviously you want to have more than one customer, and every customer might have their own preferences or maybe even their own proprietary algorithms on how they prefer to do it. Do you go in and sell a shrink-wrapped solution, "Here's our solution for portfolio optimization," just picking on that particular topic?

Or do you work with the customer to integrate into their systems in the way they like to think about portfolio optimization?

Roman: Well, actually, we do both things. We have an optimization engine, which is the core of our technique.

It works with quantum technology, and it's actually hybrid. It's combining quantum and also quantum-inspired technology, and this is why it's working so well. But every client is like a little baby. So at the beginning, we typically work with the client directly on a given project, and then we understand exactly the specifics of what they need.

Because even for such a specific problem as portfolio optimization, you may find lots of differences between one bank and another, because they may apply different restrictions to the problem and so on and so forth, and this makes the mathematical formulation of the problem different.

So we can tailor the problem to the specific needs of the bank and then run our optimization algorithm. Now, by working with different banks on different projects, we have a pretty good sense of what the most common constraints are that most of the banks, I would say 90% of the banks, are actually applying.

And we can incorporate this into our solution. In this sense, we are building a product, which is customizable up to some point, but which already integrates most of the things that the banks are requiring. And then, a posteriori, if some client needs a little bit of fine-tuning, we can also provide it.

Yuval: I spoke recently with an industry analyst about quantum adoption in the enterprise, and we spoke about who's delivering the solution. One concern that he raised, and I'm curious about your perspective, is that enterprise companies don't trust the quantum vendors to truly understand their business.

Because a quantum vendor can say, "Well, I work in optimization, and I work in chemistry, and I work in finance, and I work in supply chain." And because of all these different things, you might be sort of a jack of all trades and a master of none. How do you respond to that? Do you think that a company like Multiverse is best positioned to deliver these solutions?

Or do you want to work with a partner that specializes in supply chain or chemistry?

Roman: I think we are best positioned, and let me explain why, because I think we have a different approach than many quantum vendors out there.

The point is the following. I have the feeling that many of the quantum companies in industry come from very abstract ideas. There are a lot of people from academia involved and so on, and it's a natural thing that they just develop solutions and then go and try to sell them to industry.

And then they face the problem that the actual client may have a very different problem from what they were expecting. This is something that, at Multiverse, right from the beginning when we started the company, was completely clear to us: the client is the first thing.

You have to understand the client. The client rules. It's the client who tells you the use case, not the other way around, and it's the client who tells you the problem, and you have to work out the solution. So from day one, we have been obsessed with the client, and it has worked really well for us. The problem that the client is proposing may not be the thing you have in mind, but that's the actual problem that they have to solve.

And obviously it may be very difficult, and you may think that maybe this is impossible, but you have to work and think very hard to reverse engineer and see how you can actually solve the problem with quantum technology. And if you manage to succeed in doing this, at the end of the day you will end up with a solution that is actually useful for the client, and the client will understand you and you will understand the client better.

But obviously this is a very hard thing to do, because it's not like, "Okay, I come here with all my wisdom on quantum technology, and I'm going to tell you what you have to do." No, this is not the way of selling anything. Maybe you built a fantastic machine, but it's completely useless because it's solving a problem that nobody cares about.

It's actually the other way around. It's the client who tells you what you have to do. And since we started with this philosophy from the beginning, it went very well. Also, when we started Multiverse, already three or three and a half years ago, we had a very strong focus on finance. We were 100% focused on finance.

That was our knowledge. And in finance, we were really very good. We had a lot of domain expertise, and it worked really well for us. And what happened is that, since we were so successful in finance, people from other verticals came to us and said, "Hey, could you do something like this? I have a problem in optimization, but in the energy vertical, or in manufacturing. Do you think you could help me?"

And obviously, if you are a startup and clients start coming to you and saying, "Could you help me with this problem? I'm going to pay you," you say yes, and you start to scale your business.

So we took this approach at Multiverse. We were very successful in finance, and after specializing and being successful in one vertical, we started exporting the solution to other verticals. And this was extremely easy to do because the engine, the core of the solutions that we are offering, was already built.

The same optimization or machine learning algorithm that we use for finance is the one that we use for energy or for manufacturing and so on. So in this sense, we didn't have to build anything new. The only thing we had to understand was the specifics of the problem, the jargon, what the relevant variables were, what the context is, and so on.

In some cases it was easier than others, but we managed to do it pretty well.

Yuval: How is the company organized? I mean, how large is the company, and is it organized by verticals? Is it organized by people who work on the engine and then others who do client project delivery?

What could you share about that?

Roman: Right now, Multiverse is a company with 80 people, 80-plus actually. Our headquarters are in San Sebastián in Spain, more or less one-third of the company is in our office in Toronto, Canada, and we also have offices in Paris and in Munich.

Now, we have different teams, obviously. We have a business team, and then we have a team for sales and grants. There is also a research team. And then there is a team which is extremely important, which is the rest of the tech team, and it essentially divides into two.

One part is for services, that is, for projects with clients. And the other one is the product team, which is dedicated to actually building the product. So that's how we are organized right now.

Yuval: We didn't speak about hardware yet. Do you find that particular types of quantum computers are better for one problem, whereas others are better for other types of problems?

Roman: Yeah, absolutely. Yeah, yeah.

Yuval: And does the customer care? Does the customer say, "Well, what are you using?" Or does the customer really just care about, "Give me a solution"?

Roman: I'm going to start with the second question. What the customer typically wants is just a solution to their problem.

They don't care if it's quantum, if it's classical, if it's AI, if it's whatever. They just want to find a better solution than what they have. If it's with quantum, it's with quantum. If it's quantum-inspired, it's quantum-inspired. They just don't really care. That's if you go to the actual business line of your client, which is the one that is more relevant here: the business lines in banks, in energy providers and companies, and so on.

These are the people who have the money, who have the power to put a solution into production, and they just want a better solution to whatever problem they have. It may also happen, however, that if you approach the innovation departments, they are more interested in exploratory projects, which are more research-related.

In that case, yes, they care about it: "Okay, let's use, I don't know, a photonic quantum computer, or let's try neutral atoms," or something like this. Now, having said this, and to be pragmatic, it's true that different types of quantum hardware adapt better to different types of problems.

And this is something that we have tested and seen for ourselves. This is the reason why at Multiverse we are what people call quantum agnostic. This means that we don't stick to one particular quantum hardware provider. Instead, we have partnerships with essentially everybody who has a quantum computer on the cloud that we can access.

This is for different reasons. First of all, there is still no standard on how to build a quantum computer. But second, and most importantly, as I told you, different types of problems are better solved with different types of quantum hardware. For instance, if you had to solve a problem that involves a quantum circuit with many quantum gates between qubits that are highly localized, let's say, then it's probably better to use an architecture that has full qubit connectivity, such as ion traps or neutral atoms, instead of, let's say, superconducting quantum circuits, because there you are limited to the connectivity of the actual topology of the circuit.

However, if you want to solve an optimization problem, maybe you want to try quantum annealers. If you want to try some other type of quantum machine learning algorithm that is variational, then superconducting qubits are probably your choice. So it works exactly like that.

Yuval: You've been with the company for almost four years, I think. The company is almost four years old?

Roman: Yeah, it's three and a half, essentially.

Yuval: What is the project that you're most proud of, of all the things that you did?

Roman: I'm proud of all my projects. They would kill me if I said something different, actually. No, I mean, obviously, I'm proud of all my projects.

Still, there are some projects that are very, very interesting. For instance, recently we've been working on two very interesting projects with Crédit Agricole Corporate and Investment Banking, also in collaboration with Pasqal in France, and this was a very satisfying project that went really well.

There were actually a couple of projects, and because we could explore a lot of quantum and quantum-inspired techniques, they ended up being very, very good. We also did a couple of projects on quantum artificial vision in manufacturing, which I'm also very happy about, because we could actually beat some benchmarks in artificial intelligence that were very hard to beat.

So, yeah, I would be lying to you if I told you that I have a preferred project. I like all of them.

Yuval: It's like asking you which of your kids you like best, right?

Roman: Exactly.

Yuval: What do customers think about IP? Do they worry that their IP is somehow going to get leaked inadvertently to a competitor?

Roman: This is something that you need to negotiate with the customer, absolutely. Obviously, serious clients care about IP. The typical story is that they try to keep their IP, which is normal, but it's always possible to reach some type of agreement with them.

Whatever you developed previously, there is nothing to do about that. That's the previous IP. And then when you go into a project, you can negotiate. Typically, you can reach agreements on co-ownership of the IP between the two institutions and so on and so forth, and that's typically how it works.

But yeah, they care a lot about IP. Absolutely.

Yuval: Some of the classic consulting organizations like Deloitte or Accenture take a customer through the discovery phase, through understanding: "Well, here's what quantum is about. Let's educate you, let's determine the use cases, and then let's move to a solution."

Is there a conflict between what Multiverse is doing and what such companies are doing, or is it a natural handoff from them to you?

Roman: I think it's fine if other companies have this approach, because the more people know about quantum mechanics and quantum computing, the better for everybody, and in particular for us.

Having said this, this is not the approach that we follow at Multiverse, because we think that from a business point of view, this is wrong. I mean, this is correct if you are running a school on quantum computing; you educate the people. But as a company, you are not the one who should educate the client.

And this is related to my previous answer. It's exactly the other way around. It's the client who is educating you. And for some reason, many people don't manage to understand this. Obviously, you know a lot about quantum mechanics, but this does not qualify you to really understand the problem that these people have.

So you really have to go down and ask them to explain to you what the actual problem is. You don't have to tell them, "Well, we have quantum computers, we can do…" Obviously you have to explain things to them, but you don't have to educate them so that they become experts who can tell you what they could do.

No, that's your job. That's your job. They come to you and say, "Hey, I have this problem. Can you do something about this?" And that's where you come in as a quantum expert, and start thinking, "Okay, how the hell am I going to solve this with quantum computing?" But it's your job, not the job of your client.

So in this sense, I think it's fine to educate people, but from a business perspective, I think it's exactly the opposite of what we should be doing.

Yuval: As we get closer to the end of our conversation, I wanted to ask you what you need from the rest of the industry. Obviously, your success also depends on where the hardware is at any given point in terms of development, the number of qubits, the noise, what the software tools are doing, what the cloud providers are doing.

Roman: Yeah.

Yuval: What help, or what would you like to see from the rest of the industry to make your job easier?

Roman: Wow. Huh, that's a tricky question. Well, obviously, the hardware is going to keep developing further and further. I think this is really important, and that's the main thing about the hardware.

We need to reach that point and have this breakthrough in order to understand how to make the technology scalable. Right now we have quantum computers with 10, 20, 30, 50, 400 qubits, maybe 1,000 next year or this year, and so on. This is perfectly fine, because we can find applications that are already useful for business with these machines.

But we would like to see quantum computers with 10 million qubits. Now, how do we do this? This is the real problem in hardware: how to jump from these NISQ devices, noisy intermediate-scale quantum devices, which are limited in capability, yet useful, as we are proving at Multiverse.

But if I had to ask Santa Claus for my wish, I want a quantum computer with, I don't know, 10,000 million qubits, for instance. So how do we do that? This is something that I would like to see from the hardware companies, but obviously this is one of the hardest problems that we have at this moment.

But that will obviously help the whole quantum computing community to move forward. From the business point of view, and perhaps more for quantum software companies, there are many quantum software companies out there focusing on different things.

But I think it would be very useful if we could see even more quantum software companies emerging. Right now there are lots of quantum software startups, and this is very nice, but perhaps also big vendors. Companies such as IBM and Microsoft have done a lot in quantum computing, but there are also other technology giants that it would be nice to see in the game.

Yuval: I wanted to end with a hypothetical question. If you could have dinner with one of the quantum greats, dead or alive, who would that person be?

Roman: One of the quantum greats? Schrödinger.

Yuval: Any particular reason?

Roman: Well, obviously I would ask him about his equation and about his cat, whether it's dead or alive. But he was also an interesting personality, so I wouldn't mind having dinner with him and asking him a couple of questions.

Yuval: Roman, how can people get in touch with you, and what kind of people would you like to hear from?

Roman: Well, it's very easy to get in touch with me. Just send me an email. You can find it easily on the internet, or I also have a Twitter account. It's pretty easy to find me. And I'm happy to hear from anybody who wants to talk about quantum computing.

Yuval: Excellent. Well, thank you so much for joining me today.

Roman: Okay. You're welcome.