Podcast with Alireza Najafi-Yazdi, CEO of Anyon Techniques

Alireza Najafi-Yazdi, founder and CEO at Anyon Techniques is interviewed by Yuval Boger. Alireza and Yuval speak about Alireza’s full-stack quantum pc firm that tackles all the things from cryogenics to qubits to software program, the best way to maintain quantum computer systems working with out attaching a technician to them, and way more.

Transcript

Yuval Boger: Hi there, Alireza, and thanks for becoming a member of me at the moment.

Alireza Najafi-Yazdi: Hello Yuval, thanks very a lot for the invitation.

Yuval: So who’re you and what do you do?

Alireza: Nicely, I’m Alireza Najafi-Yazdi. I’m the founder and CEO of Anyon Techniques, and we’re a quantum computing {hardware} firm positioned in Montreal. And we even have some satellite tv for pc presence within the Waterloo area in Canada as nicely.

Yuval: And I apologize, however I need to say I’d by no means heard a number of earlier than, however I solely heard of you very not too long ago. After which I discovered that you simply’ve been out available in the market, lively for various years. Is that by design or is it simply my ignorance?

Alireza: It’s to some extent by design. We really began in 2014 and actually formally launched an organization in January 2015. And in that regard, we’re one of many oldest corporations within the quantum computing market. I feel older than us, maybe Rigetti by a number of months and possibly D-wave by fairly a number of years. And it was round that point that John Martinez and his crew joined Google and the Google quantum effort began, and IBM additionally began their latest superconducting effort. So, we’re nearly as outdated because the trade is.

Yuval: And I feel you’re constructing an entire system, each the {hardware} and the software program and all the things in between utilizing superconducting qubits. Is that right?

Alireza: That’s right. So we really, we’re fairly distinctive in some facets, and that’s we construct each main part of a superconducting quantum pc in-house. We construct all of our personal dilution fridges, which I’ve seen shocks many individuals. They don’t count on a quantum computing {hardware} firm to have the ability to construct its personal dilution fridge. We construct our personal managed electronics. We construct and develop our personal firmware stack and we sort of cease proper on the SDK that’s after we handed over to the consumer and different companions to construct compilers or subtle algorithms and whatnot. In order that’s the place we’re primarily targeted on the more durable and the peripheral of a quantum processor.

Yuval: That feels like a significant enterprise. I feel there are very massive corporations who attempt to do full stack options. Are you able to give me a way concerning the dimension of your organization? How many individuals, or what sort of funding that make it doable to do a full stack answer?

Alireza: Positive. We’re not too massive a gaggle. We’re 25 folks, however we’ve been at it for fairly a very long time now. It’s greater than seven years nearly in December. That’s going to be eight years of engaged on this mission so we have now a nimble and really environment friendly crew, however we additionally took a while to construct all these things.

Yuval: And I feel you may have a quantum pc that truly works that you simply’ve bought a number of. Is that right?

Alireza: That’s right. So we have now a system that we constructed for DRDC. It is a Canadian protection analysis institution right here in Canada, a part of the division of the nationwide protection. And that’s a six-qubit machine. That’s the extent I can disclose. And we are actually within the strategy of constructing a second unit. It’s really a really good and industrial unit for knowledge facilities. And that’s going to be put in someday subsequent yr, hopefully the primary half of subsequent yr at Calcul Quebec, which is an excellent computing middle right here in Canada. And that serves various universities right here in Canada. So yeah, we’re within the strategy of constructing that system. And the primary section of that system, we’re going to put in a 12-qubit chip. And the system is, as I communicate to you, we’re nearly all the main parts, {hardware} parts are both on the ground or they’re being built-in. So we’re going to begin testing within the subsequent couple of months.

Yuval: What makes your system completely different or higher than different quantum computer systems and particularly different superconducting quantum computer systems? It doesn’t sound like you may have extra qubits. Are they higher qubits? Are they higher related? Are they designed for various use? Assist me get the differentiation, please.

Alireza: There are two issues, Yuval. So right here at Anyon, our actions comply with two prongs. One, we have now our personal inside R&D that’s targeted on a really explicit sort of structure and superconducting qubits and a novel qubit structure that we’re following. And the opposite stream of our work in the meanwhile focuses on delivering close to time period intermediate scale machines, NISQ machines, to knowledge facilities and space adapters.

So these two machines that I talked about, they function that NISQ architectures. They’re transmon qubits. They’re tunable transmon qubits with tunable couplers. To some extent, just like the structure that Google has adopted, we adopted the identical philosophy in lots of facets. We like that philosophy very a lot. In fact, on the management electronics stage, cryogenics, all these are important, and likewise even on the quantum processor, there are particular issues which can be very distinctive to ours.

What we have now been striving to attain is greater gate constancy and minimal crosstalk. So on the crosstalk, we have now actually a state of artwork efficiency metrics. You could possibly put the crosstalk, actually talking, to zero, and efficiency metrics by way of gate fidelities, constantly, we’re getting efficiency metrics, which you’ll name us among the many high three gamers by way of efficiency of the qubits. They usually’re getting even higher and higher as time goes on.

And scaling up the variety of qubits is not a difficulty. We will scale as much as few dozens, let’s say to 60, 70 qubit with the present platform that we’ve developed. It’s only a matter of test dimension of the consumer as a result of constructing these machines prices some huge cash and any individual has to pay for it.

Yuval: I used to be wanting on the web site and a few of your press releases. And it feels like the 2 of the shoppers which can be talked about are the federal government or defense-related prospects. I’m guessing that these prospects usually are not utilizing it on a public cloud, like a Braket or Azure. Is that by design or is that this simply occurred this manner that the primary two prospects are authorities prospects?

Alireza: A part of it occurred by design. It’s only a philosophy that we have now by way of how a eventual quantum pc could be used. We predict a quantum pc goes to be a {hardware} accelerator. It’s going to be sitting subsequent to a traditional excessive efficiency computing infrastructure. So I discussed Calcul Quebec and the supercomputing middle. So we’re very excited as a result of this machine that we’re constructing for them goes to be straight coupled into their present HPC material to Narval, which is the biggest public supercomputer right here in Canada and the 81st or so largest supercomputer on this planet.

So there’s lots of fascinating integration work being carried out between the HPC and the quantum computing infrastructure we’re constructing. And due to that, let’s say placing a machine on the cloud and having a TCP/IP API name, that’s simply gradual. However we don’t assume that’s how a machine could be ultimately used.

We perceive these are smaller-scale machines, maybe largely for training and attempting proof of idea, however however, you need to actually transfer towards the fitting path. In order that’s why we didn’t go straight into the cloud. And there’s additionally with regards to cloud, the enterprise case has not been there, or not less than I’m personally not satisfied concerning the cloud but. Constructing a machine prices hundreds of thousands and hundreds of thousands of {dollars} of capital. And then you definately put it on the cloud and also you cost the consumer a number of cents a shot. And I’m unsure for those who get return on funding anytime quickly. And on the finish of the day, we’re a industrial firm. Unit economic system issues. And that’s why it was to this point has not been commercially enticing sufficient but.

Yuval: How about uptime? After I have a look at computer systems on the cloud, they’re not up 24 hours a day. They’ve restricted home windows of operation. If you ship a pc to a buyer. Do you need to ship a technician that tweaks the qubits day by day? Or how are these computer systems maintained?

Alireza: It’s an excellent query. And this is without doubt one of the issues that has been the topic of serious actions right here at Anyon. We’ve been engaged on growing automated calibration programs that, as you mentioned, tweaking. Sure, you want some calibration. You want common calibration, maybe 24 hours and even shorter time intervals. And also you don’t desire a technician subsequent to the machine. You need this to be carried out robotically. So there are lots of quantum management ideas. Software program engineer ideas have gone into constructing the infrastructure to be sure that these items may be maintained with out direct interference by sys admins or our technicians.

Significantly this one that’s going to the supercomputing middle. There’s lots of fascinating necessities by way of sustaining it. The uptime ought to be 24/7 for lengthy, lengthy intervals of time. So it’s so far, we’ve run our machines for month and month with no drawback. Usually, we have now to simply heat as much as swap a brand new era of chips. So we’ve been capable of preserve these machines up for fairly a while. And we’ll proceed to watch and study from that have.

Yuval: Given that you’re a full stack firm, do you want any assist from some other trade participant? I imply, for those who had been controlling the quantum computing trade, what would you may have folks try this they’re not doing at the moment or do that will help you transfer quicker?

Alireza: Nicely, we’re not positively controlling the trade, however we had been sort of controlling our destiny and our know-how. And that’s been at all times the concept, however there’s lots of room for collaboration. As I discussed to you, we’re so targeted on the {hardware} that we strategically have determined to let others care for algorithms, maybe compiler optimization, and issues of that nature. And I feel this can be a nice space of collaboration. Benchmarking is one other nice space of collaboration. Those that have explicit experience in benchmarking. They usually need to go from one {hardware} to a different. They’re greater than welcome to speak to us. And we’d love to listen to from them as nicely. That’s one other space of collaboration. And there may be in between lots of parts that both we don’t make, or we don’t need to make anymore, that we like to see provide chain for.

A great instance is a dilution fridge. For instance, dilution fridge. Once we began in 2015, there have been simply two corporations on the time that you can name them industrial. One was Blue Pressure, and the opposite one was Oxford Devices. And we weren’t positive in the event that they had been purchased out by our opponents, however with these large giants, what could be our destiny? So we determined on the time it makes each the strategic sense, and likewise for long run, if you wish to go bigger and bigger variety of qubits and construct larger and greater programs to have our personal dilution for programs, we’re most likely going to maintain a few of these very key gear or key parts internally and construct them internally. However we’re at all times looking out to see what others can do and take some load off our shoulders.

Yuval: And by way of functions, do you’re feeling that your computer systems are greatest for one explicit sort of software like optimization or chemistry or one thing else? Or actually is all the spectrum for you?

Alireza: We construct what’s known as common quantum computer systems. These are gate-based machines. And in idea, you may run any algorithm you need. You’re simply restricted by the coherence time of the qubits, and the gate facility is similar as you had been working on, let’s say, IBM or Google’s machine. That being mentioned, a very good query is, what’s one of the best software for a quantum pc?

And there’s additionally some corporations who’re following application-specific quantum processes or architectures and an fascinating dialogue is what’s it, what precisely is that? And the way would that play out in the long run? So in the meanwhile, we consider it’s good to stay as generic as doable. So we’re going to proceed engaged on constructing gate-based common machines and attempt to make the {hardware} extra correct, and dip our toes into fault tolerance and error correction. However that is, I feel, a really lively subject of analysis and nonetheless, everyone’s at very, very early levels of this.

Yuval: Alireza, how can folks get in contact with you to study extra about your work?

Alireza: They will attain out to us by way of LinkedIn, by way of our web site, by way of Twitter. They will, in the event that they need to speak to me in particular person, I’m each on LinkedIn and on Twitter and I’m responsive.

Yuval: Wonderful. Nicely, thanks a lot for becoming a member of me at the moment.

Alireza: Thanks very a lot, Yuval.

Yuval Boger is an government working on the intersection of quantum know-how and enterprise. Often called the “Superposition Man” in addition to the unique “Qubit Man,” he may be reached on LinkedIn or at this e-mail.

October 9, 2022