FirstQFM is solving quantum computing's hardest scaling and performance bottlenecks across the quantum stack. CEO and co-founder Vish Ramakrishnan explains why developing a model that understands the quantum machine itself will unlock the next generation of practical quantum systems, and why the Nordic region is an unexpected frontier for deep quantum innovation.
What is FirstQFM and what are you building?
We're building machine learning foundation models that solve quantum computing's most critical bottlenecks. We started with the calibration problem for superconducting quantum hardware. At the time of our incubation, calibration was performed manually or with the assistance of narrow machine learning models, which were trained to understand the problem, but not the device. We took a different approach and started with the harder problem of training a model to understand a device and its modality, and then applied this model to the calibration problem.
You've been an entrepreneur since 18-what drew you to quantum?
I've always been driven by hard problems that require deep technical understanding. Earlier ventures taught me that survival in tech means constantly learning and adapting. Quantum is the ultimate convergence: it's physics, computer science, AI, and infrastructure all at once. When you see a fundamental bottleneck that nobody else is solving at scale, and you have a path forward, you can't ignore it.
What's the core problem FirstQFM solves?
We work with quantum hardware developers to overcome their critical performance and scaling bottlenecks. We started on the control layer for superconducting devices. Since quantum systems drift over time, calibration must be performed daily. A manual approach can take 3 hours for a 100-qubit superconducting device. At scale-when you have a million qubits instead of 100-traditional methods become impossible. Our approach enables scaling by embedding information about the device and its modality, minimizing the amount of device time and measurement information needed to calibrate a quantum computer. We are currently extending our approach to problems across the quantum stack and to other modalities.
How did you arrive at this solution?
We started developing machine learning foundation models for quantum computing prior to the release of ChatGPT and the ensuing generative AI boom. At the time, narrow machine learning models were being explored as a means of calibrating quantum hardware. We had the fundamental insight early on to take the concept of foundation models and extend it to the quantum computing domain. We could see that applied machine learning in many domains, including text and image generation, was drifting in that direction and we wanted to get there first for quantum computing.
The breakthrough came from working out how to train a foundation model for an industry leading superconducting device and then applying it to the calibration problem. We immediately realized we were on to something category-defining and followed up with patent applications. While many were skeptical initially, we now see giants, such as NVIDIA, publishing articles in Nature on the necessity of AI-for-quantum and the potential of foundation models for quantum computing.
What's your earliest proof of concept?
We tested on one of the largest quantum hardware systems available and achieved a breakthrough reduction in gate execution time-the time needed to perform an elementary operation on a quantum computer-while also improving fidelity. The longer it takes to execute your operations, the more the quantum state decoheres, leaving you with noise. A reduction in gate execution time means you can execute more operations and minimize idle time.
Why does coherence time matter so much?
Quantum systems are inherently noisy. Every gate operation you're executing is a race against decoherence. And each gate is performed with some chance of error. Our compression of execution time and improvement in fidelity buys you breathing room-more operations, more accuracy, less loss. Lowering error rates below critical thresholds is also necessary for implementing quantum error correction codes.
How is FirstQFM different from other quantum optimization efforts?
During the time of our incubation, work on AI-for-quantum concentrated on the development of task-specific machine learning models. Such models attempt to solve a single problem with limited context about the domain and a narrow, specialized training set. Analogously, natural language processing was once a fragmented field with many specialized models for solving problems such as sentiment classification and text summarization. It is now self-evident that most of those problems are best solved by a foundation model that embeds deep context about the text domain, such as an LLM. This is what we’re doing, but for the quantum computing domain.
What happens at scale-why does this matter for the next 5 years?
Today, hardware developers are calibrating 100 qubits manually every 24 hours. Tomorrow, systems will have a million qubits. You cannot manually calibrate a million qubits with existing approaches and many of the alternatives require even more device time. You need something that minimizes device time and makes use of deep context about the domain and modality. That's the most promising path to production ready quantum systems. Our approach to improving device calibration isn’t just fast enough to use in production settings; it's the architecture that allows quantum to actually be useful in industry.
What's your take on quantum computing's hype cycle?
Quantum is real, but we're in the phase where hardware capability is racing ahead of software maturity. Everyone wants to run algorithms on quantum machines, but the machines themselves are fragile, noisy, and hard to operate at scale. The unsexy truth: the companies that win in quantum will be the ones who solve the operational challenges first-calibration, error correction, reliability. That's infrastructure work. It's boring but critical.
If you're a quantum hardware vendor-why FirstQFM?
Because we partner with you to solve your most challenging performance and scaling bottlenecks. Your problems become our problems. We improve your gate fidelity. We enable you to operate at higher qubit counts. That translates directly to more uptime, faster iteration, better customer outcomes. We're not competing with you; we're making you viable at scale.
Why build this from Stockholm instead of Silicon Valley?
The Nordics have deep expertise in machine learning, physics, mathematics, and systems thinking. We also have the advantage of not being caught in Valley incumbency-we're not constrained by existing quantum narratives. We can be contrarian. And honestly, the talent density for this kind of deep technical work is exceptional here.
How is it working with Luminar Ventures?
Luminar understood the problem from day one- they get deep tech, they get that quantum requires patient capital and technical founders who won't compromise on the science. What's been invaluable is their network across the Nordic ecosystem and beyond. They've connected us with quantum hardware vendors, strategic advisors, and other portfolio companies solving adjacent infrastructure problems. But more than that, they've given us the freedom to build for the long term. Quantum isn't a two-year story, it's a five-to-ten-year journey. Having investors who believe in the vision and understand the timeline is everything. They're not asking us to pivot to hype; they're asking us to execute on the roadmap.
What excites you most right now as CEO?
Watching the roadmap become reality. Six months ago, the breakthrough execution speedup felt like a research result. Now it's becoming a product. We're working on the calibration problem, building out applications across the quantum stack, and extending to other modalities. The fact that our methods aren't tied to one hardware vendor means we can become foundational. That's the vision.
One line that captures FirstQFM's mission?
To improve the performance and scalability of quantum computers through the use of machine learning foundation models.
If you're building quantum infrastructure or considering quantum deployment for your organization, reach out to the Luminar team.For more Founder Series interviews and insights from the Nordic startup ecosystem, visit Luminar Ventures' News & Insights.