Compared to a couple years ago, we’re spoiled by how many real quantum computers we can access online today. Some systems are available through multi-hardware cloud sites like Microsoft Azure Quantum and Amazon Braket; some through individual system manufacturer’s own sites. As more machines come online, and more businesses compete for quantum compute time, things are starting to get a little complicated. Strangeworks is working on making it easy for customers to control access and costs, all while providing shorter queue times to specific systems. Sometimes the best quantum computer for your use case is the one you can actually get some time on!
Guest Speaker: Cesar Rodriguez, Chief Scientific Officer at Strangeworks
The Post-Quantum World on Apple Podcasts.
Quantum computing capabilities are exploding, causing disruption and opportunities, but many technology and business leaders don’t understand the impact quantum will have on their business. Protiviti is helping organizations get post-quantum ready. In our bi-weekly podcast series, The Post-Quantum World, Protiviti Associate Director and host Konstantinos Karagiannis is joined by quantum computing experts to discuss hot topics in quantum computing, including the business impact, benefits and threats of this exciting new capability.
It’s hard to believe we have about three dozen quantum computers we can access online today. Some systems are available through new additions to cloud sites you know — namely, Microsoft Azure Quantum and Amazon Braket — some through system manufacturer’s own sites. Is there a way to make it easy for customers to control access and costs, all while providing shorter wait times to specific systems? Sometimes, the best quantum computer for your use case is the one you can actually get some shots on. Let’s get in the queue on this episode of The Post-Quantum World.
I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services at Protiviti, where we’re helping companies prepare for the benefits and threats of this exploding field. I hope you’ll join each episode as we explore the technology and business impacts of this post-quantum era.
Our guest today is the chief science officer at Strangeworks. I love that title, chief science officer. It feels like I’m talking to someone from Starfleet or something, so that’s always fun. I’d like to welcome Cesar Rodriguez to the show.
As you can hear from my accent, I’m originally from Puerto Rico, and I was a computer engineering student then, and that was 20 years ago. I said, “I want to program quantum computers,” and everybody and my professors told me I was very crazy about it because it didn’t exist. I’m like, “Yes, that’s why I want to do it. I want to help build that industry.” I’m very passionate about the history of computing, and this felt like a second chance to do computing again, just very different.
I went way in deep and got my Ph.D. in physics from the University of Texas, and then I worked at Harvard and the Max Planck Institutes and several research institutions until finally, I was pulled out of academia to make this thing I was dreaming about as an academic and make it happen as part of this novel industry.
We consider ourselves a software startup that happens to be in quantum computing. We’re not doing quantum computing programming to solve quantum problems. We are doing classical software tools to help those people that do want to do the quantum programming.
So, whurley is our CEO. He’s a skilled entrepreneur, very experienced at successful companies before with a core team. They know each other quite well. They’ve been in other previous startups and with exits and so on. They decided to enter this industry, and that’s how they dragged me in through a common friend. They dragged me out of academia to help them out navigate the waters here.
What we’re trying to do is a lot of the problems to use quantum computers right now. There are a lot of problems on the technology quantum technology itself because it’s so early, and I’m happy to expand on that. But a lot of them, even for the researchers like me that — I can still use it even though the computers are still very early right now — it’s all about the actual classical software problems. There are a lot of connectivity issues — there’s a lot of how-do-you-actually-make-things-work issues.
So what we’re trying to do is provide a broad solution to solve that problem to allow people to have access to all the quantum hardware out there, to make it easier for them with contracting, with connectivity, with all that, such that people can go from this stage where we’re at: People are experimenting with what can you do with quantum computing, and they’re progressing toward a stage more like evaluating different technologies and where they are.
From then, what we want to help people do is go to an exploration where they start finding what’s the value that they can get from it, to finally, of course, when the industry matures, to be like exploitation. People are doing quantum computing because it saves them time or saves them money. That’s when the industry has matured and you don’t have to explain why quantum computing is viable. You just say, “It will save you time and money.”
And in the industry, of course, it’s this early stage. It’s very early, the tech, which is more experimentation/evaluation. This is the stage we’re at, and we want to help the industry move forward after this process to a mature industry.
When you mentioned some of the things you guys do — at a high level, you’re helping people get to the machines — when you hear something like that, it sounds a little bit like one of the cloud environments we’re all used to, like Microsoft or AWS. How would you compare, let’s say, to AWS Braket? What would make accessing machines through you guys a little different?
That is one component. That is the essential component to get things rolling. But our value, it goes beyond that, and the value, it’s basically — everybody says that the quantum is very early, which is true, and there are a lot of options, which is true. We’re evaluating them all. But also just the problems of, how do you get the most out of each hardware? For that, you need to use software and need to program the software. And everybody says, “You have the software stack. When you have the hardware, then you have low-level controls and postcontrol, then maybe some different compilers and a couple of other languages, and finally, application level and visualization.”
This thing does not exist completely as a whole. There are some more mature ones — for example, AWS, Qiskit and so on. But as part of people evaluating technology, not only they’re evaluating different hardware, they’re also evaluating different compilers, they’re also evaluating different ways to minimize the errors for the hardware, and so on and so forth.
So it’s really not just the access to the hardware, but it’s the whole software stack. It’s, how do you have a lot of options so you can mix-and-match and improve things for the specific case that you want? That’s a big difference — we’re really focusing on the stability of the quantum software stack so that you can make the most out of the hardware, of course, to evaluate the hardware.
Let’s say, for example, like with Braket, that SDK doesn’t have Qiskit, but through you, we would be able to use that, right?
Yes. In fact, we have what we call a community version, which is at Quantumcomputing.com, and it’s intended for educational purposes, for enthusiasts and so on to get a scope of things. But you can go there right now, and you can see all the software that we support with all these platforms. We support almost everybody in terms of hardware and software that’s available. And the ones that we don’t support yet, we’re working toward. You can go right now. It’s free. It is intended as like for me to play with, so you can go and run stuff on real hardware right now, and we have free hardware. We have some free hardware from IBM, for example, that is only for us. We have, of course, all the free hardware that’s out there, and, of course, the paid hardware that’s out there as well.
So, as convenient as cloud providers are, I do every once a while notice, “Oh, this is missing,” because, of course, I’ve seen all the machines and access in all different ways. So, for example, with Rigetti, if you were to go directly Rigetti, you have all this extra level of control, and IBM does similar things. But I don’t see quite that level of control in, let’s say, something like Braket. Are you guys giving that full, like, right down to the lowest level?
Yes. We give the option for people if they want to use Rigetti direct or Braket. Braket has a lot of advantage, too, like the billing and so on. We want to give people the options so they can access Rigetti through one or through the other.
For example, with IBM, they have opened a lot of the exclusives, more like direct hardware-control stuff. Now they’re pushing heavily on what they call runtimes, which is a better-paired classical-quantum computer, so you can cycle better in optimization and optimize algorithm better. They launched it with us before they launched it officially internally for IBM.
If you want to go do some quantum optimization algorithms, something like that, you can go to our platform and start running that right now, just test things out. In some ways, it’s a bit easier even than IBM to just run some basic stuff. Of course, there’s still a lot of in-depth stuff that you have to do, you can choose which way you want to.
But the convenience of having all these different ways to access hardware with all the software platforms, it’s intended that you can have a place where you can have your own code and run it. This is when you intend to run the same code several times, let’s say, to evaluate different hardware. Maybe you want to evaluate them in time and across the different optionalities so that you know, “This seems to be improving for this use case” and so on.
So on top of that, what we’re working toward is to provide some basic benchmarking tools — just help the users evaluate their own code and evaluate the different options that they have as well so they can improve on it and make strategic choices.
Yes, that’s important, because certain types of hardware and hybrid solutions are the ones that we’re going to see performance boost first in. For example, D-Wave’s hybrid solver, I believe that’s going to be the first quantum advantage we see. So, it’s important to be able to access everything if you want to test the real effectiveness for your code. So, you can code first and then select the targets after, is that correct?
Yes. We make it easier with the different hardware. With hardware, you can compare among them or comparing simulators and so on. Now, we just really want to make easy the whole workflow of, you run, you decide if you need to run more stuff, you decide if you need to run somewhere else, and a lot of this is really comparing live what’s happening — or as live as possible. We are trying to help this iterative process of evaluating the hardware, trying to squeeze a little more out of your code, and that’s what our goal is.
Do you do any internal performance testing to pass on to users? With Qiskit runtime, they had some pretty bold claims at IBM about what kind of boost that would give to performance.
We don’t. We want to be the Switzerland of quantum. We want to be very neutral, so we have access to, as you said, IBM and AWS. And we have every machine that’s online, we have our users, too, and we have very good relationships with all the vendors. And of course, each machine has its own very nice things or what it’s good at, and there are things that, maybe they struggle a little more. And this is healthy for the industry. It’s very healthy that users compare them, and they start pushing them in different directions. I don’t think the industry is going to go forward just from each individual provider coming up with one metric that they’re very good at. What matters is having users testing them and seeing what they’re more useful for.
In that sense, we don’t see value in us telling people, “Oh, this machine is the best given this metric,” because the metrics will change, metrics can be gained, and ultimately, what matters is usage. We do want to build tools so that the users can decide for themselves what’s more important, and ultimately, the usage is the one, the only boat that matters in the race. We don’t want to call winners, but we will organize the race and have everybody race and have the users cheer.
Because there’s that community version, are you also using the other sense of the word community? Are people sharing experiences, like, “When I tried this type of optimization, I found that this hardware worked better,” or are you going to be building any kind of crowdsourced feedback like that for users, or is that just still in the distance/future?
Right now, the community version, it’s really intended for enthusiasts, and a lot of these more hardcore tools for benchmarking and so on use the more advanced tools, and this is part of our SDK, which is more for private use. This is more of a prototype thing, the same access and so on, and these are the extra features that we think are valuable for people to make. This is how you make strategic decisions. This is really the value we have. It’s not the passthrough with the hardware. It’s making it easier for people to try it all in meaningful ways and learn from that and improve their code within meaningful ways to them, like, “This is what we’re trying to do.”
We see that not so much a part of the community feature, but definitely more of the features of a big enterprise wants to try some use cases, they want to compare around, they want to improve it a little internally, but it is a real pain in the neck for them to set up everything for this one use case, and so we try to make it easy for them.
And the SDK, when you first launched, it was an online tool — you just go to it and log in?
Yes. That’s the community version.
Are you now working on a local SDK?
Yes. We have users already. We’re still in beta, but we have users from different enterprises and scientific users as well.
Was there a demand for the local SDK for any specific reason? Was it for concerns about privacy, like not giving away too much of what one particular customer is trying, or did they just want to run it on obscure hardware?
That’s a good question. We find that there are different kinds of users. There are some people that think, “I just want to test this one thing. I want to figure out what it is.” That’s the community version: “I want to see what it is. OK. That’s interesting.”
But of course, when you have quantum computing, ultimately, it’s a processor for scientific workflows, like quantum computers, scientific computing. That’s ultimately the role of it, so it has to fit in within the way people like to use it. So a lot of this, of course, is local.
So one is like, the scientists want to use the visualizations, the standard stuff, make it tightly integrated with the existing way they do things. The other one is, some enterprises specifically said they wanted this, and some of it is geopolitics: Quantum computing is, of course, very international, and there is a lot of international competition between different sectors. And sometimes they specifically say, “No, we want to keep it in this country,” and there’s an easy way to help them do that. And other ones were security. There are some companies that care a lot about security toward their property or other security aspects, like, “We don’t want some other country to spy on us,” things like that, or “We don’t want anybody else,” so they did want this locally for some things.
Of course, right now, with the industry, there still has to be some connectivity to the outside. But the idea was, at least they can do some things internally and then try to limit what they pass through and try to own the code as much as possible.
You mentioned IBM a few times, and Qiskit. Are you planning any other kind of non-IBM technology that’s coming? Any other machines? Just like now, you have this access to the seven-qubit machine that’s prioritized, do you anticipate any other kind of specialized access coming?
Yes, we do that. I cannot make the announcement yet, but we’ll make it very soon. But the idea is, it’s similar to a model that we have with IBM with special access to make some things freely available for some kinds of users that are chosen in collaboration with how we’re providing and so on, such that we can give to our providers more feedback and so that the users also get some special perks. As you said, the queuing times are terrible.
We do have this in the works — I cannot make a concrete announcement yet. But the idea is really to keep expanding this. We are friends with everybody, and we want our offering to show that we’re supporting everybody and that we want users to try them all. We really do. We think this is healthy for everybody, and so we want to be the ones pushing that.
Yes, it’s exciting. And, of course, sometimes you record a few weeks before something airs, so to the listeners, you might want to check out the show notes for Strangeworks’ site because, by the time this airs, there might be more announcements. You might want to see what this machine actually is.
In general, do you find that customers want to pay to shortcut those queues? Are they asking for ways around the wait times, or are they finding them to be tolerable?
Everybody hates the wait times. These are the researchers and the corporation and so on. The researchers, the people that actually do the programming, what they want to do is always shorten the wait time.
Yes, definitely. I know I want to do that. Before we slip away from this beta idea, was there another beta program coming that you wanted to talk about? I don’t know. Because I’ve heard about one — I have some insider knowledge — something called Backstage Pass? I don’t know if you want to explain that.
Just making sure it was all the same.
I wasn’t sure if I was allowed to say the name.
Are there any ways to then compare the results of what you get, like reporting, or extrapolating a feature like performance trends?
We’re big on extrapolating here. We love to do the charts and say, “When we get this many more qubits, we can expect this much more performance.” Is there anything like that?
Yes. I’m very excited that you asked that, because with some of the industry friends, I will announce very soon. That’s exactly what we’re doing. What we want is people that have specific use cases to be able to do some very basic benchmarking, running on different hardware, make some strategic decisions about, how good were the answers, how fast did they run? And, of course, this is very nuanced — it’s a lot of work that you do toward that.
But there is another layer, which you just alluded to, which is, we have a lot of hardware. You can compare right now all the hardware — that’s good. But in six months, there’s going to be a lot more hardware online, and we have to support even more stuff, and can you compare course times? And as long as we can have a trade of “How is this performance? This specific use case, how is it performing?” On top of that, some of the big enterprises, they have many use cases they’re interested in, and you can group them together to get a trend for this area of interest and see how they’re doing and make this kind of analysis.
Once you do that, of course you can make also broad industry trends. If this area belonged to these industries, this seems to be a little harder for people. And this helps industries themselves, enterprises and industries. It helps enterprises to decide — everybody wants to know, “When should I start in quantum? When is that inflection point that happens? Where is the crossing point?” I call that a point. There are a lot of names for it, but I like to call it quantum value, which is when you can explain that you do quantum because it saves you time and money, and so everybody wants to know what that is. And the truth is, nobody knows, and there’s nothing wrong with that. If somebody tells you they know when that happens, then don’t trust them, is what I would say.
However, of course, the enterprises need to prepare for this point, and there’s a lot of value right now in this uncertainty of helping them navigate that so they can make their own strategic decisions. They can grow the teams as they see fit. They can focus on the hardware that they find more useful through these reports and these trends that are moving forward. That’s our value: We are very honest, and we recognize that there’s a lot of uncertainty in the industry, and that means that we can help people out, navigate this, because there is a lot of value to be made.
Because you guys are framework-agnostic, hardware-agnostic, it looks like you could just focus more on managing the whole experience. Because there are costs of hardware in the back end, and it’s only going to get more complicated, honestly, as more and more companies — this reminds me of when the cloud was getting big. There was this concept of the cloud of clouds — you would have some company manage all your cloud accounts and all your access. I almost feel like this is like a cloud of quantum clouds.
That is definitely a critical aspect of it, because we found that none of the other things make sense. That is an essential part of our offering, but we’re trying to go beyond that — not only give them access to all of them but also help them decide where should they put their money in.
And part of the cloud of clouds, a lot of it is like, it’s not really the difficulties of writing the right line of code to make them work. A lot of it is the business side of making the deal and buying the hardware, and each company has different ways they do it and different rules and such, and we have to treat everybody different, and they change the queueing systems every now and then. There are a lot of those traditional business problems, engineering problems, that we try to manage so that we do it once and then all the users benefit, as opposed to each single programmer having to pull their hair trying to make this happen.
Would you be able to then use a buying-in-bulk power strategy in the future? Would you be able to buy a bunch of time and divvy it up for smaller users?
It depends on the hardware provider. They’re all very different, and they have different policies on this. However, I would say that because of the good relationships we have, we have been able to convince some hardware providers to sell some smaller pieces than they normally do through us. They do understand the reality that this helps the users, of course, and it helps the hardware provider as well.
To make a big investment to buy a lot of compute time, it’s something that’s in the cards, for sure. Of course, that’s a big plate. It’s a really big plate for the start of our stage, but it’s something that we definitely keep in mind.
It sounds like you’re doing a lot of the business facilitating on the back end.
Yes, and that’s going to become increasingly important. Costs can get out of hand really quickly here. I’ve said numerous times, we always get a little nervous when we’re running certain POCs, because you try to predict the cost of a shot, even, or how many shots it will take, and things get a little scary. Honeywell, they have a great tool that predicts that. Do you guys pass along any APIs to something like that to the customer? Any kind of estimators? Do you have any of that you create, or do you just pass along whatever a company does to estimate how much it will cost to run?
There are different areas one can estimate. They started to do some basic stuff, and it’s very difficult to make a good guess. We have many of them, and we allow people to mix-and-match what they see fit with the understanding that we cannot make that prediction, but we can help them make their own predictions to make smart choices about this. Do I want to — it’s regarding some extra precision digits. Is it worth my money, for example, or is it worth my time or such things?
There are things that we really believe in, while we also recognize that there is some unpredictability. However, there are a lot of opportunities just on cushioning some of this, operating and realizing the fluctuations for the users. And these are things that, depending on user feedback, we might make some of the decisions about: Do we want it this way or that way to help them out with that? We’re definitely at a stage where we’re having our first enterprise users and getting the feedback to see which areas of this they will benefit the most from.
Yes, it would be interesting if that kind information can be shared between users without revealing anything sensitive: “The last thousand shots we ran of this type of problem, each one cost approximately this on these machines,” and then you could start having those available to people so they could see. That might be useful.
Of course, in helping everybody do all this experimentation, there’s a lot of very valuable data that we have of the whole industry and where things are going, what is useful, what people do use, how do they use the hardware and so on. These are things that we definitely want to use to help users even do better. This one you have, it’s like one example of that, but this is exactly where we’re directed toward.
You have accounts, obviously, that you create for Strangeworks. People have credit and whatever that they’re using in your account. They can be billed directly that way to use a machine, or can they also bill to their cloud accounts, like if they have some volume agreement or something — let’s say Microsoft?
Yes. If they already have access to the machines in some way, they can use their keys, and it works great. And if they need help getting that, we can help them as well.
That sounds like a good approach. Do you have any kind of throttling in case you want to have some division of like trainer/trainees or something and you’re worried that they’re going to run up a huge bill because they don’t know what they’re doing?
Yes. We call this team tools, and they make sense for big institutions, of course. As you mentioned, this is like a classical problem. There’s nothing quantum about that. But the management of the credits, the sharing and so on, if you are researchers, you don’t have so much money. You want to make sure the grants — don’t burn it all up or something. We think there’s a lot of value in that, and we do have team-management features. Also, at the institutional level, for when a big institution has a lot of research — a big university or a group of universities — they have many researchers in that. How do they allocate? The university might buy the credits. How to help them to best allocate it properly? This is a lot of those tools that we have to make it easier for individual license.
I guess the best way to get started with something like this would be for people to go to the community version and see what it’s all about?
Yes — I was going to say someone had to nab that URL eventually, and you guys took it first, so congratulations on that one. It’s just so obvious. And, of course, in the show notes, we’ll mention that and your company site, too
Thanks for giving us this insight. The business side is going to be impacted heavily by this new cloud-of-cloud approach, so it’s exciting to see what you guys are doing. Thanks again for coming on.
Thanks for having me over.
We’re adding a new short segment here called “Coherence: The Quantum Executive Summary.” I just want to take a moment to highlight some of the business impacts we discussed today, in case things got too nerdy at times. Let’s recap.
There are about three dozen quantum computers available right now. Some are accessible by multiple online services, and depending on how you choose to access these machines, you’re given different levels of programming control.
Getting the most out of the quantum software stack can be tricky. Strangeworks is a company looking to solve traditional business challenges around using emerging quantum technology. A few years ago, companies started managing multiple cloud accounts for customers to create a cloud-of-clouds approach. Strangeworks is doing the same for quantum access. Unlike some other quantum companies, Strangeworks is more engineering-heavy than physics-heavy. They’re focusing on facilitating business and relationships between customers and providers.
As quantum computers get more powerful and time on them becomes even more precious, this approach of managing access may become crucial rather than a nice-to-have. We may even see benefits from what Strangeworks learned by working with more customers and providers —perhaps better efficiencies or ways of estimating costs — and their throttling feature sounds pretty handy if you have a department of experimenters that may get a little too curious for a particular quarter’s budget. You can learn more by visiting the links in the show notes and starting with the community edition.
That does it for this episode. Thanks to Cesar Rodriguez for joining today to discuss Strangeworks, and thank you for listening. If you enjoyed the show, please subscribe to Protiviti’s The Post-Quantum World, and leave a review to help others find us. Be sure to follow me on Twitter and Instagram @KonstantHacker. You’ll find links there to what we’re doing in Quantum Computing Services at Protiviti. You can also DM me questions or suggestions for what you’d like to hear on the show. For more information on our quantum services, check out Protiviti.com, or follow @ProtivitiTech on Twitter and LinkedIn. Until next time, be kind, and stay quantum curious.