Kanav Setia, the co-founder and CEO of QBraid, is interviewed by Yuval Boger. Qbraid is a one-stop platform that allows users to access various quantum software and hardware solutions. Kanav and Yuval discuss the benefits of QBraid for both beginners and advanced users, the company’s plans for future development, the potential for high-level programming languages and API-based solutions, and much more.
Full Transcript
Yuval Boger: Hello Kanav and thank you for joining me today.
Kanav Setia; Yeah, it’s my pleasure. Thanks for the invite.
Yuval: So who are you and what do you do?
Kanav: My name is Kanav Setia. I’m a co-founder and the CEO of QBraid, a quantum computing company. A little bit about my background. I grew up in India and got my bachelor’s in aerospace engineering from Indian Institute Institute of Space Science and Technology It was a college by the Department of Space in India. So after my graduation, I worked with the Indian Space Agency for four years and then came to Dartmouth for a Ph.D. in physics. During my Ph.D., I got a chance to work with both the Google and the IBM teams.I contributed to both OpenFermion and Qiskit Chemistry.I interned at IBM twice and towards the end of my PhD, I decided to start a quantum computing company.
Yuval: And what does the company do?
Kanav: qBraid is a one-stop platform that allows people to access all the different quantum software and hardware. We provide access to software and hardware from more than 10 different companies. That includes IBM, AWS, Microsoft, Google, and of course, QuEra.
Yuval: So if I want to write in Qiskit, can I do that with your environment? And why would I want to use your environment relative to going to the, say, the native environment of individual providers?
Kanav: That’s a great question. The reason you would want to use Qbraid is because everything’s already prepackaged. Even with Qiskit, you may have to download Qiskit, pip install, and maintain all the different environments. Whereas on Qbraid, all of that is pre-built for you. Maybe you can use IBM’s cloud service for Qiskit, but we go even further than IBM’s native cloud service provides you access to one version of Qiskit at a time. That is the latest release that they have. Whereas on qBraid, you can get access to multiple releases, so that way you can have one Python environment working with the latest release, one Python environment with the old release, one for GPUs.We now also support GPUs, which means if you want to use Qiskit with NVIDIA GPUs, you can do that on Qbraid.
Yuval: Would the benefit be primarily for beginners to say, “Hey, instead of having to learn how to install the packages and maintain them, I just go into Qbraid, sign up for an account, and I can get going in a few minutes”? Or do you also see value for more advanced developers?
Kanav: That’s again an amazing question. So for beginners, the value is clear. For the advanced user, if you gather the work that goes into maintaining each component, it adds up. And essentially what we want to do is we want to remove all of that work by automating it. So if you think about it, maintaining Python version is one task, then maintaining different Python environments on top of different Python you may have, that’s another task. And after that comes maintaining the right version of Jupyter notebooks, maintaining the right version of each and every package you have. And then if you want to access various different quantum hardware, you have to configure each of those categories. And then that comes with cloud access, right? So here we’re talking about not only the sysadmin work that allows you to configure Pythonic issues, but also DevOps that allows you to configure all your credentials correctly so you can access various quantum computers for your workflows. And on qBraid, we manage all of that for the end user.
Yuval: How does that scale if I have a team of developers that wants to adhere to certain corporate policies about updates or code style or budgets, how much I can submit, and so on. I understand how it would work for a single developer, but how would that scale?
Kanav: So it scales quite well for organizations as there is minimal upkeep required from their end. A company can make an organization on QBraid and they can buy qBraid credits, and disburse those credits to their developers, control what feature each developer can access. Further, they can control how much compute developers can get and then they can control how much credits each developer has allotted for connecting to various different quantum hardware. Most of these features are already available on qBraid, but we are working on some features that will allow for even finer control such as controlling what quantum computers are accessible to which developers. But you can easily transfer the credits to developers, or transfer the credits back to organization. All of those things are available even today and any organization can sign up and buy credits, buy compute. So all of the good stuff.
Yuval: When you mention credits, is it more expensive to go through a QBraid? Is it cheaper because maybe you buy in bulk. How does pricing compare?
Kanav: You basically answered my question. So that’s essentially the model. We offer the same price that you would find at any of the cloud providers, and we work on our end to negotiate a better deal for us so that whatever discount we end up getting, that we keep as our margin.
Yuval: I understand the value on the DevOps and keeping the environment up to date and a quick start. Does QBraid, does the company offer additional services or additional value for developers?
Kanav: Absolutely. And this is where like we have an interesting story as a company. We built this incredible platform, everyone liked it. And then we went on to try to sell it to the enterprise world. And that’s where we realized there aren’t enough quantum software engineers who can use quantum computing in their workflow right now. There’s a huge demand and need to educate and bring developers at big companies, up to speed so that they can start using quantum computing in their workplace. So what we do is, as an introductory tiered service, we also collaborate with companies to do workshops on QBraid where we teach them all the different verticals that they might be interested in. That includes quantum machine learning, quantum chemistry, quantum optimization, and teach them the basics of quantum computing and guide them on how they can introduce all of those things in their own work.
Yuval: Some people may say that they don’t care what’s under the hood, but they just want a solution. They want an optimization API or a machine learning API or a chemistry API. Is that something that you offer? Is that something that is perhaps in your plans?
Kanav: It’s definitely in our plans. Unfortunately, the field of quantum computing is not where it needs to be, where you can provide a single API that people can just call with their problem and get the solution. We totally understand that as an end user, you may want a service like ChatGPT, where you can send a text token as a query and receive an answer. Essentially you don’t want to train the model yourself and that’s understandable. If that’s the requirement, then people may have to wait a few more years before we make that happen as the quantum hardware will need to catch up. But then there are many companies which want to capture more value than what can be captured by using an API. Those are the people we’re trying to go to right now.
Yuval: Give me a sense of your users, if you can. Do they come from a particular industry? Are they particularly interested in one type of application? What are you seeing in that regard?
Kanav: So currently we have more than 8,500 registered users who have come from various programs. We’ve supported various quantum computing hackathons around the world. We’ve been consistently at MIT’s IQ Hacks since their beginning. So we get a lot of users from such hackathons. And then there are a lot of professors running various quantum computing courses. We also do a lot of enterprise workshops or these demos where even industry folks join in. So I would say a decent number of our users are beginners and who are starting to explore this new field, but then also there are like a lot of industry folks who come from machine learning backgrounds or quantum chemistry who are starting to learn quantum computing. And we’re beginning to see a good number of users this year who are becoming pro users. These are the users who know quantum computing enough and they like to use quantum computers and access various quantum software and hardware. And they’re using qBraid because we’re solving a big problem where they do not have to maintain a lot of these environments, software and hardware connections themselves.
Yuval: What can you tell me about usage statistics? Are people using more simulators than physical computers? For computers, which is most popular?
Kanav: Definitely more simulators than the hardware access. Qiskit still remains one of the most popular packages. Almost everybody starts their journey with Qiskit. If I had to take a guess, it’s like 80 to 90% are still Qiskit users. We have a close partnership with Amazon as well, and we’ve tried to promote Braket. So there’s a good number of Braket users and Braket’s gaining traction slowly and steady. A big plus point with Braket is also that it also supports many different quantum computers. So we’ve been able to work with it and build hardware connections quite readily. And so we have Rigetti, IonQ, QuEra, and many other quantum computers courtesy of Braket. There’s this really cool demo I do all the time where if I have a room full of people, say 25 to 50 people, I walk them through creating a QBraid account, connecting to quantum computers from Rigetti, IonQ, and QuEra, and running their first quantum program, submitting the first job, all within 10 minutes. And if I remember correctly, I did that for you too.
Yuval: Yes you did. I think it was probably shorter than 10 minutes. Good for you. Tell me a little bit about the company. How large are you? Are you funded, where are you located?
Kanav: So currently we are six to seven people, six full-time, one part-time, and then we have an awesome developer team working in India supporting our web development-related tasks. So in total I would say like around 11 and then we always have a few interns working, coming in and leaving. So yeah, around 11 or 12 people in total. And we raised $1.9 million so far through angel investors, venture funds, and we just got awarded up to $4 million from Q4Bio, the Welcome Leap grant that just recently came out. The phase one is $1.3 million, which is what we should be receiving in the coming months. And then if we perform well, we go on to phase two and phase three and could unlock another $750K to $2.7 million in total. And then this morning we received another amazing news that another grant we submitted in collaboration with NCA&T came through, which is like a million dollar per year grant for up to five million and we are one of the subawards in that.
Yuval: Congratulations. Professionally speaking, what keeps you up at night?
Kanav: What keeps me up? That’s a great question. I would say there’s not a whole lot of things that keep me up night. I’ve come to peace with things that are under your control and things that aren’t. Maybe I can comment on things that I think about and have some concerns. As a quantum software company, we can only control software. So my concern is if the hardware companies are going to deliver on their promises. Many of these companies we closely work with and we see how much hard work they are putting and we are quite bullish on their progress. So if hardware gets much better, it will be an amazing news for the entire industry. If we can make these awesome quantum-computers with fault-tolerant qubits that work as promised, then software is relativel easy to build and could be built at an amazing pace. I believe we could be addressing the needs of the end users if the hardware comes through. On the other hand, if the quantum hardware progress slows, tough times may follow.
Yuval: You mentioned that Qiskit is the most popular environment for programming. I mean, under your system. Do you think it’s high level enough? Do you think that as people go to larger and larger machines with more qubits, with error correction and so on, do you think there’s gonna be a need for something else or do you just see Qiskit evolving and continuing to be the environment of choice for most people?
Kanav: Absolutely, there’s going to be a need for high-level programming languages. It is reasonable to say that 15 years from now, to solve a problem people will not have to build the quantum circuits themselves. That is not how end users will be using quantum computers. It might be API-based or certain other interface where you just call a few commands, submit your problem and get the solution to your problem. Qiskit, so far has evolved with the needs of a quantum software engineer and it will be interesting to see if they keep doing that by developing high level packages. An interesting question that I ponder on is when is the point when a high level programming language could be built? When does that jump happen? At what point can you can say that my underlying hardware and software work as promised and it allows me to depend on it and I can build my higher-level applications. And that point is not super clear to me because ultimately we (qBraid) would want to do that as well because you get to capture a whole lot of value if you go that route, like build a high-level computing service where people can just call an API and get the answers to their problem. And as soon as it becomes clear, we will also be starting development for some products in that category.
Yuval: Moving from development to production, let’s assume that I created an algorithm using your environment and I’m happy with it, and now I just want to run it in production. Do I still continue to use your environment or do I take that code elsewhere?
Kanav: If you can define what is meant by production environment, that would help me a little bit. Like, well, what does deploying in production mean in this context?
Yuval: Let’s say that I need to run an optimization problem every 15 minutes. Every 15 minutes, I get a new set of data. I want to run an optimization problem. Let’s assume that the computer I want to run it on is actually available. Should I be using your environment, or should I take that code elsewhere?
Kanav: I think if you are at a stage where you want to deploy your solution that you’re happy with, we will help you make that happen. We have thought about this problem from the moment we started out. Our motto is, learn, build and deploy. And again, it comes to that critical question, when do we think people will find enough value that they would want to deploy? Constantly, as we built the infrastructure for learn and build, we’ve had this question in the back of our mind, always and consistently. So we’ve worked with a few companies where we build custom environments for them where they can do this deploying part. And meanwhile, we’re constantly learning if we had to do this at scale, what we would need to do. Imagine three years from now there’s this sound programmer working in their basement and they stumble upon the most amazing code that solves real problems. How do you empower that developer to just call a few functions that packages their code, turns it into an API, and makes it available for the rest of the world and help that developer monetize their quantum solution. Basically how do you build an app store that works for quantum? That’s basically what Apple did with iPhone, right? So the question for us has constantly been, when is the right time to build an app store? You build it too early, you don’t get enough traction. You wait too long and then somebody else has captured the market
So, constantly, we’ve been thinking. It has been a hard question to answer.
Yuval: And as we come to the end of our conversation, I wanted to ask you a hypothetical. If you could have dinner with one of the quantum greats, dead or alive, who would that person be?
Kanav: Feynman for sure. That’s a no brainer. Richard Feynman, I think I just get such a boost out of listening to the very few videos that he has online. The disrespect he has for the establishment and the way he gets things done, I just admire that so much. And I think that also comes with a price. He might be a little bit arrogant and mean for being forced to go for a dinner with someone he doesn’t know, but I think it’ll be worth it.
Yuval: Excellent. Kanav, thank you so much for joining me today.
Kanav: My pleasure, Yuval. Thank you for inviting me to your podcast.