IEEE Quantum Podcast Series: Episode 18


Portait of Helmut KatzgraberHelmut Katzgraber

Global Practice Lead, Amazon Quantum Solutions Lab


 Listen to Episode 18 (MP3, 21MB)



Part of the IEEE Quantum Podcast Series


Episode Transcript:


Brian Walker: Welcome to the IEEE Quantum Podcast Series, an IEEE Future Directions Digital Studio Production. This podcast series informs on the landscape of the quantum ecosystem and highlights projects and activities on quantum technologies. Today we're speaking with Helmut Katzgraber, Global Practice Lead at the Amazon Quantum Solutions Lab. Helmut shares details on his company's structure and his views on the state of quantum as it relates to real world solution scenarios. Helmut, thank you for contributing to the IEEE Quantum Podcast Series. To get started, can you share a little information about yourself?

Helmut Katzgraber: Sure. Hi, everybody. My name is Helmut Katzgraber. I'm Global Practice Lead for Quantum in AWS's Professional Services. I am a physicist by training. I've spent most of my life in academia working on organization problems. Later, I went more towards a quantum computing area and after doing some consulting gigs, I joined Microsoft for 3 years and now I'm with AWS leading a team of interdisciplinary scientists that solves really hard problems for our customers.

Brian Walker: So Helmut, can you just provide a little overview on your company's structure?

Helmut Katzgraber: Absolutely. So AWS is really investing in the long-term when it comes to quantum computing. And our program is focused on four main pillars. On one hand, we have Amazon Braket. It is a managed quantum service that is in full GA where you can access quantum devices made by third party hardware providers and at some point, down the road also our own devices. We have the AWS Center for Quantum Networking. This is something relatively new that we launched recently where we focus on developing software, hardware related to quantum networking. Initially, for example, the focus in quantum key distributions for quantum secure communications and then we have our Center for Quantum Computing, where we are building our own qubits based on superconducting systems. The reason why we are focusing on super connecting systems is because this area has the most expertise and is the longest established one. Although these qubits come with a twist that allow us to do more effective for our corrections schemes that are being developed on the theory side of our Center for Quantum Computing. And then next to these three pillars, there is the Quantum Solutions Lab that I lead. We are part of Professional Services, and we are the customer facing branch of Amazon's Quantum Efforts. So we work hand-in-hand with AWS and new customers and our main mission is to identify real world applications where quantum computing can have an impact but also try to raise the bar with adjacent technologies like machine learning, high performance computing operations research such that we can really assess what a quantum system should be able to deliver in the future such that we can have really an advantage over classical technologies.

Brian Walker: So what differentiates the AWS approach to Quantum as it pertains to real world applications?

Helmut Katzgraber: Yeah, absolutely. Amazon has a very interesting culture that is very dear to me. We tend to say that we work backwards from the customer and then starting from the voice of our customer, from these market signals that we hear, then we go in and develop the technology that best addresses these problems. And so in some sense, you see that we are not just telling the customer, "Here, use this." We're giving the customer a choice that allows them to select the best tool to solve a given problem. And so I can give you two examples. On the one hand, we have our Braket service where you have different quantum backends and the reason why we have this is because today, we don't know yet which technology is going to at the end really be the one that scales and delivers value and impact. And so it's important to experiment with very different platforms and different qubit technologies simply because this will allow us to determine which technology works best for which use case. Similarly, in my team, the Quantum Solutions Lab, we have customers that come from all kinds of backgrounds, from what we like to call quantum curious that basically say, "Hey, we heard about quantum. What can you do with this?" to actual companies and customers that have expert teams in-house. As you might imagine, these teams hare very different needs. And right in between, lie the ones that I'm very interested in, which are those that say we have a really hard problem. We do not know how to solve it. Can quantum technologies help? And so again, we start with the problem of the customer and then we build up on that and develop a solution that either really assesses how a quantum technology could be used for such a use case down the road, but also ideally, delivers value today with a scalable solution. And then finally, the third differentiator is to my knowledge from the big quantum providers out there, we are the only ones that are offering networking solutions to our customers. At this point, still, in proof of concepts, but we're ramping up quickly as always.

Brian Walker: So Helmut, speaking of your customers, can you cite some recent examples of solutions you're working on with companies?

Helmut Katzgraber: Yeah. So our customers come, as I said, with very broad backgrounds and different industry verticals. I can give you a handful of examples in three main verticals. For example, in health care and life sciences, we took a very different approach. Whenever you hear health care life sciences and quantum, the immediate thought is, can we do quantum simulation, for example, to solve a chemical problem? The reality is that clinical trials are the largest fraction of the cost when you're developing or manufacturing a chemical or a drug, and so we latched on to that part and said, "Well, how can we leverage quantum technologies to optimize a clinical trial enrollment, for example? And this is one of the use cases that we looked at where we came up with a very clever solution that allows a customer now to basically make real time decisions on how to, for example, fill enrollment centers for a particular trial in the most effective way. In finance where we see the strongest customer signals, we have worked with different companies such as Fidelity and Goldman Sachs. With Fidelity, we're really going far fast forward and looking at very interesting potential future applications of quantum technologies. I cannot quite disclose yet what we're doing, but what I can tell you is that we have had engagements with them in the past and we are developing again new algorithms based on quantum technologies that would allow us to solve some very complicated problems in the finance sector. Similarly with Goldman Sachs, who is a company that has an expert team of quantum scientists, we actually did a very theoretical resource estimation work where we said, "Okay, if we want to solve this, say in this case, portfolio optimization problem, using a quantum machine and a second order cone programming algorithm, how many qubits-- what are the-- sorry, the circuit depths, and what are the T gate counts that we need to be able to solve the problem at scale? And well, as you might not be surprised, it takes a hell of a lot of these qubits and T counts, so it's still something that is going to be far, far out in the future. And then the final example I want to give you is in automotive, there we work with BMW to solve a complicated problem that relates to their production and that is really trying to optimize the motion overall what's in the plans. And so, when you are manufacturing a vehicle or anything, in some sense, that uses robots, you need to plan out the motion of the robot ahead of time and ideally, you want to find a solution that makes sure that the robot fulfills all quality standards in the shortest amount of time. So in the case of BMW, we were looking at spraying PVC sealant on a vehicle and the goal was to optimize the motion of the robot such that this can be done in the shortest time. And here, this is a prime example of this duo track approach which if used where on one hand, we map the problem onto a quadratic unconstrained binary optimization problem, which is the language that quantum machines understand, so that leveraging quantum and annealing technology showed that the overhead of mapping this complex robot motion problem onto this quadratic unconcerned binary optimization problem makes it very hard to solve a problem at scale without clever ideas on, for example, how to break up the problem. But in parallel, we leveraged one of our physics algorithms that we used to study physical systems, combined it with an algorithm that is used in our middle model optimization at and came up with a new solution that allows the robots to run 10 percent faster. And if you can imagine roughly what kind of volume of vehicles a company like BMW makes every day, 10 percent can be a very large number.

Brian Walker: We see a lot of hype in the media related to quantum and quantum computing. Can you share your thoughts on that subject?

Helmut Katzgraber: For me, it's always shocking when I read a story or see a presentation where somebody promises that we will cure cancer with quantum computing or where we'll solve a problem where any expert knows no, this cannot happen. It's going to be way too complicated. The constraints are too difficult. And so, one of the things that is very important to us at Amazon in general is to steer away from hype and be the voice of reason. We are very pragmatic in that we clearly state this application is not good for a quantum machine. Whereas, if we have something where we see a potential, we also clearly state, "Hey, this application might have potential." So I'd like to quote John Preskill, one of my colleagues that once said, "Quantum science and technology is exciting when discussed accurately and responsibly." And I think that accuracy is of course difficult, given the technical level and you have to simplify this for a broader audience, but responsibility is something that we all should put our focus on because it's in our responsibility to not overhype a technology that potentially might not even have an impact in a particular industry vertical.

Brian Walker: How do you see the IEEE Quantum Initiative helping to advance the technology space?

Helmut Katzgraber: So as I said at the beginning of the podcast, right, there's all these technologies that are currently being explored. Ion traps, superconducting, quantum dots, et cetera, et cetera. And we don't know yet really what the technology would be that will scale or that will be able to scale in a meaningful way. Similarly, and this is even more important, if you look at one of my colleagues, Stephen Jordan's, quantum algorithm zoo, you will see that the number of algorithms designed for quantum computers is a given exponential speedup over their classical counterparts is very small. And so, there is a hell of a lot of work to be done to A, improve the technology; B, more importantly, put a focus on error correction techniques and error correction schemes; and C, also develop the next generation of algorithms to run on these quantum machines. And so, as a scientist myself, it is super important to leverage a broad community. And so it will definitely take a whole community to move this forward and this is where I see IEEE's role in kind of coalescing all this to be a leader that brings experts worldwide together such that we can advance the technology and make the strides that we're trying to do such that we can deliver something of value sooner than later.

Brian Walker: Helmut, thank you again for taking time to speak with us today. In closing, do you have any final thoughts you'd like to share with our listeners?

Helmut Katzgraber: It's interesting. As I said, you know, there's a lot of misunderstandings relating to quantum technologies. And I think one way to look at it very pragmatically is that once we think about our quantum device as a type of coprocessor, the same way we have GPUs or FPGAs of other cutting-edge technologies to solve some high or some very difficult compute problems. In the same way that, you know, for example, a GPU is embedded in a system with a CPU that basically drives, for example, the operating system, I see quantum machines really being very valuable in a hybrid mode. And I think that it's really important that we start thinking with this mentality that a stand-alone quantum machine cannot really do much, the same way that if I give you a GPU card, you would say, "Well, what do I do with this paperweight?" We need to really start thinking about how to integrate in classical and quantum in developing the next generation solutions. And so my team in general, which is made up of Operations researchers, physicists, high performance programmers, machine learning specialists, we love really hard challenging problems, and we don't shy away from them; on the contrary, we steer towards them. So one of the things that I think is very important is that we think first about the problem and then work from there to a solution. And if the solution involves quantum technologies, great. And if not, we need to be honest to ourselves and to the community that there's better methods we can use to solve a problem.

Brian Walker: Thank you for listening to our interview with Helmut Katzgraber. To learn more about the IEEE Quantum Initiative, please visit our web portal at