Daniel Porat and Prof. Emanuele Dalla-Torre of Quantimize join Yuval Boger to discuss their work in quantum optimization and hybrid quantum-classical algorithms. Daniel, CEO of Quantimize, describes how the company focuses on real-world optimization problems in industries like energy, e-commerce, and scheduling, leveraging quantum computing to enhance efficiency. Emanuele, a professor at Bar-Ilan University, explains the company’s core algorithm, ECHO (Efficient Correlated Optimization), which significantly reduces qubit requirements for solving complex combinatorial problems. They highlight their commitment to benchmarking quantum approaches against classical solutions and their efforts to provide accessible quantum computing tools to non-experts. The discussion also covers the evolving quantum landscape in Israel, the potential for commercial quantum advantage today, and the balance between publishing research and maintaining competitive business strategies, and much more.
Transcript
Yuval Boger: Hello, Daniel. Hello, Emanuele. Thank you so much for joining me today.
Daniel Porat: Hi, Yuval. Thank you.
Yuval: So, Daniel, who are you and what do you do?
Daniel: I’m currently managing QuantyMize. In the last two years or so, we started to establish this team of experts from quantum computing, AI, machine learning, and industry experts. Our main goal is to build algorithms, access, and quantum advantage for quantum computing. These are our main goals. We’ve been working for about two years now and I believe we have a very nice achievement developing algorithms and looking into the industry needs and have some very nice POCs and other achievements.
Yuval: Emanuele, who are you and what do you do?
Emanuele Dalla Torre: My name is Emanuele Dalla Torre. I’m a professor of physics at Bar-Ilan University near Tel Aviv. I’ve been working in the theory of quantum many-body physics since my PhD. And recently with the advent of quantum computers, I found them very interesting as a topic of research and a couple of years ago Daniel came to me and said, “Well, do you know how to use quantum computers? Maybe you can teach me and solve all optimization problems at once.” And that’s how QuantyMize was started. We haven’t solved all the problems yet, but we are making very nice advances.
Yuval: Is there a particular type of problem or problems that you’re focusing on to begin with?
Daniel: Yeah, I believe so. So that’s a good question because this is what we asked ourselves in the first months. Can quantum computing really solve all of the optimization problems or what kind of problems can we solve? And we end up with four guidelines. What kind of quantum computing optimization problems can we solve? And based on these guidelines, we started to develop our technology. It’s a very important process that we went through because starting from the beginning, we understand that now, especially in the current status of quantum computing, we cannot solve every optimization and we should focus on the ones that can be solved in the near term. And that’s what we end up with, these four guidelines.
Emanuele: Yeah, I think it’s also important to focus on real-world high-value applications because if you solve an optimization problem very efficiently, but that optimization problem doesn’t bring any value, then your work is not very tangible. So that’s why we actually came up with specific sectors that we would like to focus on, at least at the beginning. And that’s where we have developed our first proof of concepts, where we applied the algorithms that we developed to solve real-life problems, especially in the fields of energy for the problem of power grid optimization, where we showed the prototype that has up to 50% efficiency gain in implementation.
And we’re also looking at scheduling, which we find to be an exciting new frontier in time management based on quantum technologies, where we cooperate with the leading companies in the sector. And we also have other collaborations with commercial partners.
Yuval: I think I saw one of your white papers and it used quantum annealer. Are your algorithms designed to work in the future on gate model machines as well, or are you focusing on the annealing technology?
Emanuele: So the big challenge is really to take the real-life problem and translate it to a model, for instance, the QUBO model that is applicable to quantum computers. Now the specific way that that model is then solved, whether it is an annealer, whether it is QAOA, or even an efficient simulator, that’s not the focus of our company. I mean, we are quite opportunistic. We’re taking advantage of all these technological advancements in the solving of the QUBO problem itself, but then we are bridging that gap, which is taking the QUBO and translating it to a real-life problem that has value. So the ultimate technology that is used is not very important, and what we are doing, I would say, is generic to all those technologies.
Yuval: And Daniel, I’m curious, how large is the company? I think you mentioned it’s about two years old. How many people, where is the company located?
Daniel: So we are located in Israel, Tel Aviv mainly. We believe that this ecosystem in Israel is fantastic for this kind of company. We have 10 hardware companies developing quantum computers and facilities and stuff, and other software companies. The ecosystem here is good because we have very simple access to potential customers and POCs and everybody knows everyone. So the ecosystem here is excellent. Our team is built from experts, as I said, from quantum computing and machine learning. Part of our solution is based on taking the data from our customers, doing some machine learning kind of processes, and after that, feeding it into our algorithms. So we have AI and machine learning experts, and part of our solution is to understand the problem. So for every problem that we are trying to solve, we have an expert or experts. So it’s very important for us to understand the problem in detail and when we bring some kind of solution to our customers, it’s a real life problem solution, not an academic one.
Yuval: Tell me a little bit about the business model, if you could. Some companies do more of a project by project basis. They come to a company, they come to a customer and try to solve a particular problem. Others sometimes offer an API for, “Oh, give me a scheduling problem and we’ll give you a solution.” How does QuantyMize prefer to work?
Daniel: Yeah, so if you remember our three objectives were algorithms, access and quantum advantage. Access is part of our technology, an important part. Our vision is that any kind of data analyst or non-quantum expert that would like to run some kind of optimization problem or any kind of problem in quantum computing could do that. So with our platform that we are building, we’re going to give access to these kinds of guys that will have the ability to run optimization problems and other problems directly in quantum computing without any kind of middleware. This is our main business product, providing the platform, algorithms and access directly to quantum computing.
Yuval: There’s a famous tech CEO that said that quantum is 20 years away. How soon do you think customers can get true commercial value besides the research interest from quantum computers?
Daniel: We can tell you that our current prototypes and POCs show that we can add value to our customers today. Now we’re not saying that it’s like 100x value, you know, the quantum computing vision, but using a hybrid type of solution, meaning intelligently using quantum and classical advantages, we can add value today in a specific problem, as we said, that we can show that this advantage is valid. And we expect in the next year, this year actually, to have even revenues based on this technology, and that means that we are adding value today to our customer.
Yuval: Emanuele, what can you tell me about the algorithm? Has it been published? Is it about efficient encoding of a QUBO problem or is it broader than that?
Emanuele: Yeah, thanks. So actually, there is a fight between Daniel and myself because after all, I’m a professor of physics and I want to publish everything I do. You know, every time I do the smallest thing I want to publish, and Daniel is much more business-oriented. By the way, this is actually one of the strengths of our company, that we have people coming from really different backgrounds and that really helps us thinking out of the box. So that’s why I like to work with Daniel and with all the other people in the team. But so no, we haven’t published it. It’s there. I mean, if I could share my screen, you would see the paper already. You know, we’re almost at the submit button, but we haven’t submitted it. Our main structural algorithm, we called it ECO, efficient correlated optimization, allows us to solve a specific combinatorial problem called the multi-constrained knapsack problem, which is actually a model that can be helpful for solving a broad range of combinatorial problems.
So what we showed in our paper, in our unpublished paper, is that, but you can see some of the results in the white paper, is that we can indeed efficiently map the problem to the hardware. So if we focus, let’s say, on the current D-Wave hardware that has 5,000 physical qubits, we can solve a problem that has up to 100 items and 200 constraints in a situation where the conventional mapping would require 800,000 physical qubits, which are, of course, unavailable. So that’s one example.
Daniel: So I can add that our focus when we are building this algorithm is our four main objectives. One, minimizing qubit requirements, streamlining quantum circuits, meaning decreasing the number of layers in quantum algorithms, optimizing for real hardware, meaning replacing complex hardware-challenging operations with more efficient native operations, and using hybrid solutions. These are the four main objectives when we are building our algorithms, and we find it very very useful, especially in the current very challenging ecosystem of this technology.
Yuval: How generic is the algorithm? I mean, you mentioned a certain type of problem, but when you go to every customer, do you have to redo the algorithm to match their particular constraints, or is it as simple as “oh, here’s a QUBO formulation, now make it better”?
Emanuele: I would say that it’s midway between the two options that you gave. What we are doing is to solve a few specific classes of problems that can be mapped efficiently. I cannot claim that we will have a panacea that will solve every problem at once, but what I can say is that given a specific type of problem like scheduling, we know how to map them efficiently, or the knapsack problem I mentioned before, that’s another example. And in fact, if you look at the community of optimization, they have classes of problems, and if you know how to address each of them efficiently, or at least many of them efficiently, that will bring value to the companies.
By the way, we also combine quantum optimization with classical optimization, so if you come to us, you will also get the best classical solutions, and that’s very important for us, because we all the time benchmark what we do with classical solutions. That’s actually a critique I would have for many other companies, where they show a promising quantum result, and compare it with a random solution. No, actually you should compare it with a good classical solution and show that you are at least as good as that, and that’s what we do as a normal routine.
Yuval: Daniel, maybe you could share some information about the POCs that you did, I mean even if you cannot disclose the specific name of the customer, maybe the type of customer and the type of problems that you were able to address?
Daniel: Yeah, sure. So as Emanuele said, our basic algorithm is ECO, it is a structural one, and based on ECO we had an international kind of cooperation to implement it in an e-commerce kind of solutions, meaning you have the question in e-commerce, one of the questions is, optimization question is what kind of menu you’re going to show your customer based on hundreds or thousands of parameters, and so this is an NP-hard kind of problem and what we do, we implemented this kind of ECO solution to this problem. Other areas that we are working on the energy, power grid energy. In recent years, renewable energies are coming in, a lot of them, all kinds of energies and there is a kind of battery that you need to charge and recharge sometimes and it’s hard problem, and you have of course the limitation of the lines, and so in general it is a very, very hard problem, optimization problem. It’s involving a lot of money and we successfully developed a solution, a prototype we can call it right now, that we can show that, you know, 50% of some of the parameters can be saved with this kind of solution. Scheduling, we are into this too, and we are doing services too. And R&D collaboration on security with industry on the clustering problem.
Yuval: I think you said that you’ve been CEO for about two years and I don’t think you were in the quantum field before. How is the quantum market different than what you expected it to be now that you’re two years into it? What can you tell us about the difference between your expectation and reality?
Daniel: Yeah, so yeah, I’ve been at the edge of technology almost all my career. In 1999 I was at Intel, I worked at Intel for about 12 years, and in 1999 my manager asked me to implement some kind of technology called the internet into the operation of the company. And nobody wants to work with me because nobody believed in that, it was too slow too, I don’t know, whatever. So I’ve been at the edge of technologies all my career at Intel and later on at a company called Liola Technologies that I managed for about 10 years. They did optimization for semiconductors. So optimization and implementing technologies for optimization was most of my career. And now I understand that quantum computing is the next technology to solve these kinds of problems because with classical computing there are not good solutions for really hard problems. So coming into the quantum area was very, very interesting, especially working with Emanuele and his team. You know, it’s always been a very nice process until now and hopefully it will continue.
Yuval: And as we get close to the end of our conversation, I wanted to ask you a hypothetical question. If you could have dinner with one of the quantum greats, dead or alive, who would that person be? Maybe Emanuele, you go first and then Daniel.
Emanuele: I think definitely Richard Feynman, whom I value for his capability of getting people interested and passionate about what they do. I haven’t met him in reality, but I’ve seen movies of him where he was already aged, but so passionate about the topic and you could just listen to him forever. So something about communication, I think in today’s world is so important and that’s something I always try to improve. I mean, I think I’m okay with solving equations, but how to explain them and how to let people understand their importance, that’s something I think is very important even today.
Yuval: And Daniel, how about you?
Daniel: As you said, I’m not coming from the quantum academy, but I would love to have dinner with some of the visionary guys from the companies that are building these kinds of solutions or hardware, like at Google or IBM or IonQ, and understanding what kind of process they’re looking for, timeline and stuff like that, and technology. This would be my dinner of preference.
Emanuele: I think Daniel’s hope is easier to realize. You could just arrange a dinner.
Daniel: Yeah, we’ll do that. I mean, hopefully.
Yuval: Very good. Emanuele, Daniel, thank you so much for joining me today.
Daniel: Thank you. Thank you very much.
Emanuele: Thank you, Yuval. Thank you for your interest in QuantyMize.