Nir Minerbi, co-founder and CEO of Classiq, is interviewed by Yuval Boger. Nir and Yuval discuss the challenges of developing large-scale circuits, training the next generation of quantum scientists, Classiq’s work with Rolls Royce, Nir’s view of the financial climate for quantum computing, and much more.

Transcript

Yuval Boger: Hello, Nir. And thank you for joining me today.

Nir Minerbi: Hey, Yuval. It’s fantastic to be here.

Yuval: It’s great to have you. So who are you, and what do you do?

Nir: So my name is Nir Minerbi. I live in Israel in Tel Aviv, a physicist by training. And for the past three years, I’m a co-founder and the CEO at Classiq, a quantum software company. And we are pretty busy in making quantum software scalable, reachable for everyone, and an industrial tool rather than a scientific toy.

Yuval: Could you elaborate on that? What does scalable and reachable mean for Classiq?

Nir: Sure. So one of the fun things about quantum computing is, on the one hand, there is a lot of momentum and many large players working on every aspect of the stack, software, and hardware. But it’s very early, and many of the things that we take for granted in the classical stack are not available in the quantum stack.

So basically, when you want to develop software for quantum computers these days, you are limited. You are limited to developing software at the gate level. You really need to specify what quantum gate and building blocks to apply on each qubit. And that’s pretty much impossible, both obviously for non-quantum experts but also for quantum experts.

It would be very challenging to find someone that can really develop software for more than 20, 30, 40 qubits, while industrial quantum software eventually will be on hundreds, thousands, millions of qubits. So what Classiq brings to this world is automation, design automation methods, and tools that were developed in the classical stack for 60 years.

This is what we bring to the quantum stack. And by bringing these methods of operating systems and compilers and high-level modeling, we allow quantum software to be much closer to classical software, more abstract, more scalable. It means that when you let automation do the hard work of gate-level design, so you can easily design circuits for many, many qubits.

We do that all the time. And also, this is really interesting, it allows you to be much, much more optimized than could be achieved otherwise in a manual way. So this is, in a sense, what we do. But eventually, this is a software platform that allows users to develop quantum software and applications. This is what we do.

Yuval: Many people today that develop quantum software use Qiskit. So do you see Qiskit at a different layer? Do you see Qiskit going away if Classiq continues to catch fire?

Nir: Yeah, great question. So Qiskit is a very good platform for gate-level design. It’s Python, it’s convenient. And if you know pretty well what quantum gates to apply on each qubit and you know how to design the circuit, Qiskit is probably the best platform to do that.

The problem is it’s pretty much not scalable. It would be impossible to design complex circuits with Qiskit. And also, by the way, non-complex circuits in an optimized way. Because when you do that in these levels of abstraction, you need to take care of many design considerations like connectivity of the qubit and uncomputation of some blocks and so on and so on and so on.

And we are not replacing Qiskit, we are just one layer on top of Qiskit. Actually, Qiskit is one of our outputs, obviously, when we integrate on top of IBM and on top of other quantum clouds. So I see it more as the natural maturing of the quantum stack rather than replacing Qiskit.

Yuval: Abstraction layers always sound good, but people sometimes worry that you lose efficiency because the more you try to make something very general, the less and less it becomes hardware-specific. And so we see, for instance, hardware vendors today that try to be full-stack vendors.

They say, “We wrote chemistry software that works particularly well on our hardware.” Does that mean that Classiq could be used effectively today? Or are you more looking into the future and saying, “Well, in two years, when there’ll be more qubits and more machines, this is where you really see the benefit of using the Classiq platform?”

Nir: Sure. So, of course, the product, like I think everything in quantum computing, is dedicated for the next generations of quantum computers that could actually bring value. So the product was designed for creating complex circuits with many qubits with deep functionality.

On the other hand, and this is really the fun thing when we work with customers today, is pretty much in every use case, and I can give several examples, the platform, the automation really brings much better results, more compact circuits, fewer qubits compared to Qiskit and compared to any other software platform.

And this is because when you start from the model level, you leave automation much larger room for optimization. With Qiskit and with other software platforms, you have compilation. Compilation is automatic, you will never compile circuits manually. But the compiler is very limited because you give it a quantum circuit, it’s unaware of its functionality.

So, of course, if there are two H gates in a row, it will know what it should do with it. But if a qubit is an auxiliary qubit or not, if a block is used for this and that functionalities, it’s unaware. And when you create a model with the Classiq platform, so the synthesis engine is aware of what your functional desires and also what are the hardware constraints. So automatically, it can really generate much, much better results.

Yuval: Who are the target users for the Classiq platform? And when you work with commercial customers, what does this look like? Do they just get access to the software, and that’s it? Or is there a significant service component? What does a project look like?

Nir: Great question. So basically, our target users and customers is everyone that wants to develop now or in the future applications for quantum computers. Today, obviously, there are several focus areas. One of them is working directly with enterprise customers with large banks, pharma companies, automotive and so on and so on. And there, you will find two groups. One is working with quantum teams of quantum experts. The other group is teams that have no experience in quantum computing whatsoever.

They are machine learning teams, software teams, and they are using the platform to onboard into quantum computing in the most easy and abstract way. And I’ll elaborate on that in a minute. Another sector is service providers. We work with many of them as the platform of choice. So when they educate enterprises on quantum computing projects, they will do that with our platform.

And another sector, which is very important for us is academia, both for education and for research. So these are very different sectors, but all have the very same goal to develop sophisticated, complex, optimized circuits to run them on various back-ends. We are integrated on top of AWS Braket, Azure Quantum, IBM, and also directly with IonQ and other machines. So this is an end-to-end software platform that fits the needs of most users today.

Yuval: We’ll get back to the academia in a little bit. But since you mentioned working with customers, I wanted to ask you, is there an example you can give me, maybe a customer that you’re particularly proud of the work? I don’t know if you can mention the name, but at least the type of work and what the customer was able to achieve with the Classiq platform.

Nir: Yeah, definitely. So one of the customers I really appreciate is Rolls-Royce. The Rolls-Royce team is built of quantum experts, actually. And they are dealing with a very interesting problem, which is CFD. CFD is a computational fluid dynamic simulation. Is really a problem that pushes the boundaries of classical computing to the limit.

HPCs just choke with the size of the magnitude of the problem. And obviously, quantum computers, specifically with the HHL algorithm, bring a lot of hope to solve this problem in a much more efficient way. And what this team created with the platform is pretty much something that wasn’t created so far.

And this is a full-scale implementation of the HHL algorithm from a functional model all the way to executable code. That was very, very impressive. And this is one type of customer that knows exactly what they want to achieve. They understand that creating real proficiency and assets in quantum computing is not something that you can do in two weeks of POC. It takes time.

And another type of customer, actually it’s more of a partner, HPE, also an investor. We are building together the quantum stack for HPC users. And part of the process is to onboard their team into our platform, into quantum computing. And there you see another type of users that are leading experts in their field but not in quantum computing yet.

So they use the platform in various use cases and algorithms in order to onboard into this field in the most easy and abstract way. So these are two different examples. But I think both are showing the strengths of abstraction, automation, and optimization in quantum software.

Yuval: I think that Classiq recently launched an academic version, and I think the rationale is probably people go to universities, and they learn about quantum, and then they go into the workplace. So why not catch them when they’re young and teach them how to use the Classiq platform? What does the academic edition look like? What does it mean for a student or a professor that is dealing with quantum?

Nir: Sure. So first, this is the same product, the same platform of course, that enterprise use. And what we decided to do, of course, and in this case, it’s in a deep partnership with Microsoft, is to launch a global academic program that is dedicated to bringing the most advanced quantum stack, our platform, on top of Azure Quantum and other back-ends of course to researchers and educators.

So indeed, the platform is already used in quite a few quantum computing courses teaching computer scientists, physicists, and other sectors to develop software for quantum computers. And the other sector is quite different, is using the platform for research. And there, it’s fascinating. You see so many different research aspects conveyed on the platform.

One researcher will want to develop large-scale quantum circuits, for example, the full implementation of Shor’s algorithm, which sounds familiar. We all know Shor’s algorithm. But I will be very surprised if you find someone that can actually develop executable code of this algorithm without automation.

So this is one kind of researchers. And the other kind of researchers, for example, they will want to optimize a small quantum circuit for a specific IonQ device in order to gain as much signal out of the noise. So these are really different use cases, but we support both kinds, which is nice. And, of course, for us as a company becoming the industry standard, part of that is becoming the economic standard. But it’s more than that.

I think it’s a good goal, and this is a good mean of advancing this quantum industry. I think, Yuval, you wrote several times about the shortage of talent, and we all feel that. And eventually, maybe not next year, maybe in a few years, but the world will need many, many more quantum software engineers. One way to do that is, of course, bringing more courses to the world. But the other way to do that is bringing to these people the optimal software stack for learning and researching.

Yuval: I wanted to ask for your read on the capital markets. We see that some quantum companies are laying off people. We see that the stock price of public companies has taken a dip. Some budgets seem to be moving from quantum to AI. Everything that has GPT in it is getting a lot of attention these days. How do you see it from your side, from the vendor side, the capital markets, and the interest of companies to invest in quantum?

Nir: So I think there are two types of quantum investors. And also, by the way, two types of enterprise quantum budgets. But let’s focus on investors. One type, this is the type we are working with. And this is the type that understands the real value of quantum computing, understands that it’s a long play, it’s creating a new industry, a new era in computing, and the opportunity is huge.

You can really create the next Microsoft, the next Google, the next Intel in a new computing field, but it will take time. So if you invest in a quantum company, and after two months, you ask, “What’s your ARR?” You don’t really understand the field because it’s in research; it’s not in production. So these types of investors are still super interested. We have seen some very impressive quantum investments in actually the last couple of weeks, of months.

The other type of investors are investors who are more affected by the hype. And, of course, there is hype. And I think we suffer from the hype. It’s not good for the industry, it’s not good for anyone. But obviously, this source of money is now off. And obviously, the companies that were lucky and unlucky enough to be public are feeling it more than anyone.

But I think if you take a look at a company like IonQ, obviously the share price is important. But what’s really important is they have a lot of money and a very good vision to develop quantum computers. And this is something that no one could take away from them. So I don’t think that someone should invest in quantum to see ROI this year or next year. But in order to create the next giants of computing, this is pretty much a good investment still.

Yuval: Classiq is an important player but is just one player in the industry. And there are other, of course, hardware vendors, and you mentioned HPC providers and other types of companies. What would you like to see the industry doing more of that perhaps it’s not doing enough today, in your opinion.

Nir: I think we should all do this shift, a mind shift, at least from a very early stage exploration with very vague KPIs towards production. And it really doesn’t mean that we’ll see any quantum production soon. It will take time, maybe a short time, maybe a long time. But when you create quantum readiness within the enterprise, the goal shouldn’t be to do some nice PR to design five qubit circuits and to run them on some machine and be happy about it.

The goal is to be ready. To be ready with the team, with the assets, with the IP, with the proficiency. And I think we see that more and more. And I think what I want to see, obviously… Well, it’s easy, I want to see hardware vendors moving forward toward large-scale machines with less noise and more accuracy. So that’s easy. But I think this attitude that we see more and more in the industry, this is what is really needed. And this is pretty much it, I think.

Yuval: As we get closer to the end of our conversation, I wanted to get back to the HPC, the high-performance computing part. Many algorithms are hybrid algorithms, variational algorithms, or machine learning that have a quantum piece and a classical piece. Do you think there’s a missing link, an orchestration layer, or someone that pulls together both the classical part and the quantum part? Or do you think that’s already covered by existing companies and platforms?

Nir: Definitely, there is a missing link, and we see more and more focus on it. By the way, it’s not only hybrid execution. Execution itself, many companies think or say they have integration with quantum clouds. But when you’re actually trying to build this deep integration, the ability to design a quantum circuit and to seamlessly execute it and get meaningful results via quantum cloud or indirect integration, that’s complex.

When you try to do that in a hybrid way, it gets even more complex. So I think we see more and more companies where IBM and HPE and AWS and Dell and others leaning towards this hybrid approach. And HPC is part of that. We see more and more HPC centers acquiring quantum computers, developing hybrid algorithms.

And, of course, for us as a company, that creates the standard in quantum software. So this is a great focus area, and this is the reason we are very privileged to be working with HPE. So I think there is a missing link, but there is enough time and focus to close it.

Yuval: And a hypothetical question, so if you could have dinner with one of the quantum greats, dead or alive, who would that person or those people be?

Nir: Well, I’m not very creative in general, but I won’t be very creative here as well. I would choose Feynman. I think what would interest me is to ask him. I’m sure there is going to be a very nice dinner. But what I would like to hear from him is we all like to quote him and say that the original motivation for quantum computing was simulating nature.

And, of course, we see progress there, but we also see more and more algorithms in very different fields. I would really love to dive deep with him on what he sees as the optimal way to simulate nature with quantum computers. There are many good directions, but not a very good solution or answer. So that would be really interesting.

Yuval: And last, Nir, what kind of people would you like to contact you? What kind of partners are you interested in hearing from after this podcast?

Nir: I think anyone within the industry or outside the industry that is willing to take part in this exciting field. So I hope to be reachable. And please contact me over LinkedIn, email, or anything like that. That would be my pleasure.

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

Nir: Thank you so much, Yuval.

Yuval Boger is an executive working at the intersection of quantum technology and business. Known as the “Superposition Guy” as well as the original “Qubit Guy,” he can be reached on LinkedIn or at this email.

April 10, 2023