Transcript | Application-Specific Quantum Computing – with Anastasia Marchenkova Bleximo

Squeezing every bit of performance out of a NISQ-era quantum processor is usually handled in the software stack, but what if you could actually design QPUs for specific types of use cases? Our guest, Anastasia Marchenkova, is working on this approach. We explore the possibilities and benefits of these machines, as well as the remarkable awareness work she’s been doing for years, in this episode of The Post Quantum World.

I’m your host, Konstantinos Karagiannis. I lead Quantum Computing Services in 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 will need no introduction to those in the quantum community. She’s well-known for her various roles and publications and her social media presence, where she explains quantum to the masses. She’s a research engineer at Bleximo. Anastasia Marchenkova, welcome to the show.

Thank you so much for having me.

Tell us how you got started in quantum. I know that for you, in some ways, it’s a traditional approach.

In some ways it is, but I feel like sometimes this is a disappointing story to people: “How did you get into quantum? How did you pick the field?” I went to college as a computer science major, and then I switched to biology, and physics only came later. I took a Physics I class, and I was, like, “I’m good at this — let’s look into this a little bit more,” and it always fascinated me a little bit because biology was great, but it was a lot of memorisation. I didn’t want to do that. The textbooks were a thousand pages long, so I was, like, “That sounds boring.” Computer science was great. I enjoyed it, but I wanted to use computer science as a tool more, so I moved into sciences that way, and physics was actually hard. So I had to struggle, and I enjoy that, as weird as it sounds.

After that, I started looking for research positions. I was already working in a chemical engineering lab doing transdermal drug delivery with microneedles. It was great, but the work wasn’t where I wanted to be, and part of my research of “Do I actually want to study physics?” was to go into a lab and work hands-on, and I started emailing professors. They don’t advertise research positions often on their websites. Maybe nowadays that’s changed — ten years ago, that wasn’t happening. And I emailed them, and the first professor I emailed, I think it was a biophysics lab, and they’re, like, “We want you to have taken a couple classes on this level — write back your fourth year,” and I had just finished my freshman year.

And the second lab I wrote to was a quantum optics, quantum telecommunication, lab, and they’re, like, “Great. You can solder? Awesome. Let’s bring you in. Can you come in tomorrow?” That was all I needed, and I had electronics experience from robotics back in high school, so they brought me in and I did research there for three years in the quantum space just in my undergrad, and I stuck with it. So ended up getting that degree in physics and kept going.

But you didn’t get away from large textbooks, because quantum computing books are, like, phonebook size.

They’re often large, but you don’t have to memorise that much. It’s like a couple of equations that you end up knowing. I remember my thermodynamics textbook was, like, 100 pages. It was called Mirror Thermodynamics, and it was covering a couple topics, and the aerospace engineers, theirs is like 1,000 pages of problems, and they did not like being friends with me during thermodynamics time.

We met eight years ago. It was at the 2015 RSA Conference. I cannot believe how far the industry has come since then. I was giving a talk, Quantum Keys: The Future of Encryption, and you were nice enough to come over and say hello.

I remember back in that day, I was completely different person, which you can probably attest to. And part of doing my socials has trained me that way, but I was so nervous to approach you. I’m, like, “Why would this person talk to me? I wrote one blog post about quantum and security.” And I was so nervous, and here we are. And then I got to speak at RSA recently, so it was, like, “Maybe one day, I will be in his position, speaking at RSA.”

That’s terrific. You were nice. I remember you came over. You were working at Rigetti, and it sounded so far-fetched that a startup would be building a quantum processor. In my head, I was, like, “What? How is this even possible?” and it’s not such a strange concept now.

I know. It’s exploded since then. Rigetti was one of the only companies around at the time. D-Wave had been around for many years working in various fields there, but I started a company between undergrad and grad school, and I’d done a couple years of grad school, decided startups were more fun, and at that point, how many people were in quantum and had done startups? At that point, it was me and maybe a handful of other people, so it made sense that I would join in the very early stages of this crazy market that we’re in now.

How much has your day job changed from the Rigetti days to now at Bleximo?

Not that much — well, it depends. Every day is so different, and that’s true of any startup, and I enjoy those early-stage startups. I’m not a big-company person. The biggest company I’ve worked for, I joined at about 100-something employees, and it went up to almost 1,000. And at that point, processes were already built, and there was a lot of maintenance, and I was, like, “No, let’s go back and go work for a four-person startup” or something like that.

I enjoy that day to day, everything is different — some days are very technical, and some weeks, I’m, like, “I haven’t coded in two weeks, because I’ve been doing decks and presentations and writing, and I just want to sit down and code a little bit.” And then, after a couple weeks of code, I’m, like, “I’m ready to do something a little bit different.” But it’s fun building a company from scratch. I’m grateful for the lessons I’ve learned from larger startups and seeing them grow, and working on the fundraising side, the operations, HR. There’s so much to do, and it’s a great time, but, yeah, every day is different.

Give everyone a little high-level of what Bleximo does.

Bleximo is another quantum computing company. We’re in the hardware space, but our approach is different. We’re working on application-specific quantum processors. People talk a lot about the NISQ era — the noisy intermediate-scale quantum device era, where some algorithms may actually work on smaller systems, and pretty noisy qubits. This works because the circuit depths are not super deep.

When we talk about circuit depth, we talk about how many operations you do in a row. For some of these, there are hybrid classical quantum algorithms, and you feed some part of the algorithm into the quantum computer, read it out, put it back into classical machine and do this rotation for a little while. You don’t need super-great qubits. You don’t need very long coherence times.

These algorithms that can run on these computers hopefully will show quantum advantage in big areas — chemistry-simulation optimisation. They can affect a lot of industries, but they’re not going to be able to run Shor’s algorithm, for example. That’s a very deep algorithm that requires a ton of qubits, and we’re not at that stage yet.

Bleximo’s approach is saying, “We have these qubits. A lot of companies are trying to build universal gate systems, but we are not Google. We don’t have infinite funding, and we can’t afford to spend 20–30 years to create million-qubit machines. What if we can optimise the chip itself to the algorithm?” A pharmaceutical company wants to run quantum chemistry. They don’t care if it runs a financial application as well. So you’re going to be getting these machines for very specific purposes because you’re selling to the enterprise.

That started the idea. We’ve done research, and we’ve seen that first, it makes it a lot easier to build, because if you design in a very optimal algorithm, you can say, “Qubit one is connected to qubits five and seven, so let me build those connections and optimise it, and that way, I don’t have to build every other connection.” The one thing there is quantum volume, and connectivity is talked about a lot in quantum machines. You have to have full all-to-all connectivity. We believe that, yes, that’s very important in the long term, but right now, if we can optimise to that, the connectivity needs to be exactly for the problem you want.

We’ve seen, through software and through our hardware designs, 10x improvements in the number of qubits that would be needed for certain tasks, so now we’re working on building those things. We’re also working on a lot of the peripheral systems. They are superconducting-based, and one of the things there is, you have to keep it very cold with the cryostat dilution fridges there.

We work a lot on the packaging. Our team has a lot of expertise. And the qubits themselves are not everything. The number of qubits isn’t everything, and you can increase coherence and fidelity of these chips by working on the peripherals as well. One of the things that I mention here is, Google has very low coherence times compared to IBM systems, but their engineers are so talented at squeezing out performance on the software side that it almost doesn’t even matter at that point. They can do a lot on that.

The software side, you can optimise. We work a lot on the fridges. We figured out that we took someone else’s chip and we put it in our shielding and packaging and increased coherence times 5x just by changing the shielding and packaging of the dilution fridge — no other changes.

A lot of different tracks going on as well here, but it’s an exciting time.

Are you then going to try and have other companies that make quantum computers do that — package with you for better performance?

Yes. We’ve sold a few systems like that already to academic institutions, and they’ve seen the results, seen the improvements. And we’re talking to some investors in the space that are, like, “Being an arms dealer during the quantum race is not a bad business strategy either.” There are a lot of dilution fridges out there.

We’re building our own custom system for education as well. The idea is that cooling down chips sometimes takes a long time, and once it’s there, if you don’t need to swap it out, fine. A permanent system can be expensive, but if you’re a student that wants to learn more — if you’re doing a lot of testing and iteration, which is what we’re doing — you want to have faster cool-down cycles. For educational, cheaper is better, and you don’t need more than a handful of qubits for a lot of this educational time, so we’re building a low-cost educational system. It’s not going to get as cold as the more expensive systems, but it will be cold enough to work on the quantum workforce as well. There’s a huge talent shortage in the quantum space, so that’s our approach to that angle as well.

Would that be the Rubicon machine?

Yes, the Rubicon platform.

I saw that at Q2B.

We fed our line drawings in our CAD into MidJourney to see what it would come up with and do designs on that, so that’s been a little fun experiment. People are excited about that. It seems to be picking up, and now it’s, like, “Now we have to build so many of them.”

Good choice. MidJourney is more of the sci-fi-looking output compared to DALL-E, which looks a little flat, comparatively.

It does. I feel like I’m finally getting MidJourney. It’s a feel. For a long time, I was, like, “MidJourney feels a certain way,” and then I figured out how to get around it. But it’s a fun little experiment on my side.

Yes, and that improves the general appearance. What are you planning, then? Are you going to have these chips that are going to be named by the industry, like the chemistry chip, the financial chip? Is it going to be that kind of thing, like a line for each?

Yes. We can optimise to a certain level. The idea is — and we have sold chips — once you give us a problem, we can work with the algorithms team and the hardware team to optimise to your specific problem. However, we can make a more general chemistry chip that maybe is 80% of the improvement, and our plan is to eventually put those up in the cloud once we get to a certain qubit count and say, “This is the chemistry chip, this is the optimisation chip, this is the financial chip,” and show improvements over the universal systems at this point.

That’s a good approach. So it won’t just be like selling a box to someone. You will have, like, the chemistry service — the ways to log in and get to each one.

Yes. We’re looking at it as the customer lead acquisition. If you can show good improvement even with not a fully optimised chip for companies where that extra 20% matters, they’re the ones that are going to pay for it and have the money to pay for it, and they’re going to pay to get the last 20% of performance out of there. It’s the introduction to that NISQ quantum application–specific space.

You brought up Google before. Sycamore now feels aged. It feels like that processor’s been out forever, and the very latest papers they’re putting out are still Sycamore papers. Like you said, they’re just squeezing out that little bit of approach. So you’re still refining the stack as you go in hopes of this cloudlike platform to take shape?

Yes. It’s a startup. There’s so much to do. The hardware is definitely our focus. That’s what we have our strength in, but we’re going to have to build the connections. We’re talking to folks. There’s nothing wrong with connecting to Amazon’s Braket service or SERC or any of that. That’s totally open, up in the air: Keep building, testing out control electronics, and building out those drivers is a whole thing. Yesterday, we had a discussion on that. Keep going — keep producing the hardware as well. Hardware is such a different world from software. I’m amazed now how the world actually functions with the way you have to work with machine shops, for example, to get parts done. How we buy a laptop, and how it works, is incredible to me. So, exciting journey there. Yes, the software stack is starting to be built up. There’s a lot to do.

Yes. It looks crazy when there’s a new Apple silicon chip. Now, you can imagine, once you introduce the quantum level here, what happens. What types of companies and customers are you targeting in the short term to try and keep growing?

We do have the great academic partnerships. We’re founding members of the advanced quantum test bed out in Berkeley. We work very closely with the university. We’ve sold chips to universities as well for testing with some old packaging solutions. For the actual application-specific, we’re building relationships, and it’s the typical stuff. I tell people I’m sick of this too, but simulation — oil and gas industry, pharmaceutical industry, financial industry. Those are the common suspects. They’re also the ones with money to pay.

Yes. There’s nothing wrong with reminding people that this is a real industry growing, and real potential benefits will be seen very soon. Do you have any other plans or road maps you’d like to tease here? Something you hope to see hardware-wise?

Rubicon is exciting. We launched it at Q2B, and that was our test for the market, to see what would happen, and that’s exciting to me. Again, I mentioned that the quantum workforce is something. I’ve been working on a NFT project, and I know it’s not trendy right now, but that’s another side thing. I was, like, “We’re working on the dilution fridge? We want to have it as an educational thing. What if we sell the CAD designs as NFTs so people could actually 3D-print their own dilution fridges or stuff like that and learn that way?” That’s what we do at Bleximo. One of our processes for prototyping is, we 3D-print first, and you saw the 3D-printed fridge as well there. We do that because it’s less expensive than going to a machine shop and screwing up on gold-plated hardware or copper.

I was going to say, you’re not going to print with that.

Exactly. So I was, like, “What if we do this crazy thing?” We’ll see — all my free time that I have to learn blockchain engineering and NFT stuff. We’ll see if that happens. That’s a little fun side thing.

The neat thing with Rubicon was the size. It looked like it will be very rackable one day. It looks like it could be tucked away in something.

Exactly. We’re fitting it into a standard half server rack. The idea is, a lot of these systems in the long term also will need to speak to supercomputers or be in a data center, so you plug it in at that point, and, hopefully, there are connections there to make it happen. I’ve talked to a few folks with the data centers, and it’s, like, “If we can put it next to the machines, that would significantly help in terms of adoption and have all that connected.”

It’ll make them feel like it’s a real technology they can bring in-house finally if it doesn’t need a special lab room or something. I thought that was a smart design.

It’s not as cool, I know, as the steampunk chandelier, but we’ll make it cool on video someday.

Yeah. It’s what’s inside — that’s the cool video, and then you seal it up to be practical. Obviously, you have to seal it up, or you’d be dead if you ran it.

Outside from Bleximo, you do so much with awareness. Your writing and videos have been tremendous forces for that in the industry. You’ve even given textbook recommendations not just here on the podcast, but in general, and I’m sure several folks have been inspired to join our ranks by watching your stuff, so that’s great. I love to see that. I’d love to talk about how you find time for that and how it’s enriched your interactions with the industry.

I started writing a blog back in graduate school because I kept getting the same question over and over, and I got tired of people who were having to rewrite it over and over. So I said, “I’m going to write a blog post about it.” It’s always been a part of my life. For me, writing is hard. English is my third language, so even though I’m very fluent in it, there’s still stuff I struggle with. I went to a technical school, and one of my biggest regrets now is not taking public speaking courses and not taking more writing courses, these skills that people don’t realise how important they are — communication is so important.

It became a way for me to organise my thoughts and to understand what I was learning. It was third year that I learned that people say words and they don’t know what they mean, and I would never say anything without being 100% able to defend it or know what it means. Writing was a way for me to put all my thoughts down, become more eloquent and practice those skills that I didn’t have.

I started that, and it was just a hobby. My hobbies are a little more limited. I don’t go out and party all the time, and I like being at home and waking up early, and I’m, like, “I’m going to write.” I saw so many people asking questions that I had the answers to, and I wanted to spread it out more. I grew up in an academic family. My dad is a physics Ph.D., so I knew the Ph.D. was a pathway, and I didn’t realise until I started writing and people were, like, “I wish I could go get my Ph.D., but I can’t afford it.” I’m, like, “Ph.D.’s are free.” I always knew that growing up in that academic family, but a lot of people didn’t know, so I thought, “Why not put it out there, and maybe someone finds it and gets inspired?

Just documenting — I’ve always been a big fan of YouTube, and people are, like, “You should just pick up the camera and do something,” and the big push in social actually happened to me. I always knew it was valuable because I was meeting so many so interesting and great people through just my blogs, so I knew that was valuable, and I knew that gave me opportunities that I didn’t have. I dropped out of my Ph.D., so I didn’t have that degree, but my content was out there in public, so people knew what I was talking about.

But using YouTube started because the pandemic began and people have been asking me for videos. I had two videos up on YouTube, which is that best-quantum-computing-books one and a blog from QISKIT camp out in Switzerland, I believe. That was the Switzerland one where I vlogged it just for fun, and I’d watched a video and marketing and they said, “Go on TikTok. It’s the next big thing. You can’t dance? Fine. Put the camera up and record two videos a day talking about what you know,” and I was, like, “I’m going to do this for 30 days.” I actually did it with a coworker from another job that I had. We’re, like, “We’re going to do this challenge for 30 days.” After 30 days, I had 14,000 or 16,000 followers. So I was, like, “There’s something there. People want to hear more about it.”

And I remember I actually made a friend because he saw my TikToks and was, like, “I have to know who this crazy person is, thinking they can teach quantum on TikTok.” But it helped me see that so many people were interested in learning more about it, and I had the expertise and I could bring people into that world. And then, from there, it kept spiraling, and I was, like, “People are asking me questions, but I need more than 60 seconds to explain this, so I have to make a YouTube video.”

Then YouTube took off, and it spiraled from there. As you mentioned, it was about, what do people find valuable, and what are they asking me, and I made videos to do with that, and it worked. I know it’s so stupid. People talk about, like, “How do I get into the field?” I’m, like, “Just do it.” It’s the worst advice, but that’s the same advice I got, and I thought, “It was stupid. No, I need to prep.” Just do it. Answer questions that people have about your field, and that’s the best way to start.

I’ve heard some crazy advice. It’s, like, “Just make 100 videos, and then you’ll get better.” I’m, like, “Wow. 100 videos? Sure.”

You will get better.

Of course. Yeah. You learn the storytelling and editing. Some of my favorites were ones where you take on a myth-busting angle, because I know certain things drive me crazy to have to explain over and over again. Do you want to hit us with a couple of recent myths that you wanted to smash — things that are driving you bonkers lately?

It’s almost the same myths. The TikTok algorithm is interesting. I’ll scroll, and there are people into quantum leaping, so I get that in my feed a lot. That drives me a little bit up the wall. I’m not anti all of these things. This is going to be controversial — are you ready?

I’m ready.

Astrology is an interesting framework to think about things in. The idea that Libras are better at sports, but Malcolm Gladwell, in his book, said, “Athletes tend to be born in September, because that’s the beginning of the school year.” So, if you replace “beginning of the school year” with “Libra,” it’s still works. The pattern is there.

The quantum leaping stuff is in there, and I’m, like, “It might be an interesting framework.” But I get people who are aggressive about the last Nobel Prize. That was the basis of my entire undergrad degree, and these people are coming in telling me I’m wrong. We can quantum-leap into a different universe because of the Nobel Prize, and I was, like, “I’ve been working in this field 15 years. I understand how this works.”

But the biggest myth that comes out of it is that information travels faster than the speed of light. I’ve been trying to debunk that for many years. No one listens to me, and this Nobel Prize — they’re, like, “You’re wrong. All your videos are wrong.” We’ve known about these Nobel Prize experiments for many years. That’s a big one that I keep trying to say, and it’s a very subtle point, so I understand why people struggle with it. Faster-than-light communication is a huge one, and this is where I start thinking about education.

One I’ve been thinking about for the longest time has been, we need to make quantum more accessible. Why are we calling things Hadamard gate? It’s not very intuitive. A lot of fields have a lot of words that are intuitive, and that maybe makes them harder for people to learn or understand. We use a lot of terminology that people don’t understand.

On the other hand, I was reading this book last year called Tomorrow and Tomorrow. It’s a story about this guy that cryogenically freezes his wife to find a cure for disease and then himself and goes forward all this time to try, and wakes up to try to see if there’s a cure for her illness. In one of the jumps, language has evolved so you speak math or you speak writing, and those are completely different languages. So I’m, like, “Maybe I’m thinking about this wrong. Maybe we do create a whole new language. Maybe we don’t teach quantum — it’s not intuitive, because the language structure that we currently have is not intuitive. Can we do something there?” So, a tangent, a very idle thought that I last week thought of. Maybe there’s a different approach to that education aspect.

I don’t think anyone wants to speak math — that’s for sure. Max Tegmark tried that with his mathematical universe. I don’t know if you ever saw that. That’s a pretty scary approach for some folks.

To keep on that negative note with myths and things, what are things in the industry that worry you right now?

It’s a hard time for quantum hype. I do think we’re a little bit past the peak, but last year was rough. The people that I respect always held true to their morals and their scientists, and I’m seeing a lot more people now coming back and being very truthful about where the challenges are. Sometimes, when you get marketing involved and you have to fundraise, there are a lot of promises that you can make, and people who are knowledgeable about the space will at best be disappointed, and that’s had a very negative effect on the industry. That part worries me. This is part of why I started my writing — there is very simple pop science explanation, and there are the papers, and I want to be a little bit in between and have the average technical person understand what’s going on.

I’ve seen it getting better. At Q2B, I was talking to the Quantinuum folks, and their scientists were so on point. I was so impressed, and they’re, like, “Here are the challenges that we have: We don’t know how we’re going to do this.” And in the same way, I try to be very honest in that way in my videos, like, “Yes, I’ve committed to a certain modality working on the superconducting space.” Do I think photonics still have a use? Yes and they have different challenges, so you still need a very complete picture of the quantum space to truly understand. Trapped ions, they’re, like, “Coherence times are super long,” and I’m, like, “Yes, but gate times are also longer — the shuttling time of the ions.”

“Superconducting?” That’s what we’re working on — manufacturing. There’s a reason that we hired mostly engineers at the beginning, because that was the challenge. “You need expensive cryogenic systems.” “Yes, that’s a thing. Laser cooling systems are all so expensive.” So, you have folks that are maybe giving an incomplete picture. Photonics — millions of cubes, sure, but — there’s always a “but,” because we still don’t know who’s going to get to that point.

I’ve committed to certain directions. Other people have committed to their direction, but there’s space for a lot of these modalities to change technology as a whole. Obviously, I’ve never wanted to do quantum telecommunication with superconducting qubits. That’s not what it’s good for. Number cruncher — that’s what I see it for, which is why this application-specific problem space works for it, but, of course, we’re never going to be, like, “Yeah, we’re going to do quantum telecommunication here.” It wouldn’t make sense. Even if it would raise $1 billion in funding, it’s not truthful.

So, you don’t have any significant concerns like a quantum winter or anything like that?

We’re in it already. The last four months were already pretty difficult. I’ve steeled myself already. It’s already here. We’re going through. We’re surviving. I like the idea of the arms dealer in the quantum race. I’m excited for it in some way because it’s going to force people to start commercialising technology, and that will have a positive effect when we start sharing more information between us in the end. It’s still early. We’re all still friends — and maybe we won’t be friends in a few years, but we’ll see.

Well, we definitely will be. What trends now are you most excited about? What’s your most optimistic take? What’s exciting you?

In the quantum space specifically? Trends. I don’t know — it’s not a very trendy field we’ve been working on.

Through lines you see coming to fruition — things that are approaching?

In our space, seeing that educational aspect ramp up. We never thought lasers would be such a big deal until they were. Now, lasers are used in everything, so having more people involved in the field is exciting to me. I’m excited that the U.S. is finally investing in quantum telecom technologies and quantum security, so that’s been a huge thing. I know it’s been always at the back of people’s minds, but quantum telecom — China was and still is way ahead of us on this, but there’s a company, Kinect, that’s doing great things in the U.S. I’m excited to see that spinning up and starting to run networks and lines there.

That’s awesome. I was excited to finally have you on here. It’s been a long time coming. It’s good to connect again, and I wish you lots of luck. And I know I’m excited to see what things your company does, and, of course, the things you put out on your own.

Thank you so much.

Now, it’s time for Coherence, the quantum executive summary, where I take a moment to highlight some of the business impacts we discussed today in case things got too nerdy at times. Let’s recap.

Bleximo is taking an application-specific approach to building quantum computers. Rather than just optimising the software, they’re also getting extra performance from their hardware by designing QPUs to operate more efficiently when running specific applications. Thought is given to which qubits need to be directly connected to each other, for example, to run certain gates of a specific application more efficiently. This will lead to QPUs that are suited to groups of use cases such as chemistry or finance while lowering complexity and costs. It’s a novel way of building processors that has already shown orders of magnitude of improvements in execution time with fewer qubits. The company hopes to eventually have these systems available on the cloud for ease of access by use case type. Bleximo is also making small, half-rack-sized dil fridges like the Rubicon platform, ideal for experimentation in educational settings.

Anastasia has been spreading awareness about our industry for quite a while. Be sure to read the show notes for links to Anastasia’s writing and social media.

That does it for this episode. Thanks to Anastasia Marchenkova for joining to discuss Bleximo and her personal writing and videos. You can see a video of this interview at our YouTube channel. 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 Protiviti Tech on Twitter and LinkedIn. Until next time, be kind, and stay quantum curious.