Aqarios: Shaping the Future of Quantum Algorithms for Solving Optimisation Problems

Unlocking the potential of quantum computing is not just a hardware challenge. Programming is also crucial, and there is a lack of talent that understands both programming and quantum physics.

Aqarios is building a platform to interchangeably apply classical and quantum solvers to optimization problems—including in-house developed quantum solvers and third-party solvers. It was founded on the backbone of seven years of research, and spun out from the QAR-Lab at Ludwig Maximilian University of Munich in 2021, by Claudia Linnhoff-PopienSebastian FeldThomas Gabor, and Robert Jacobi.

Aqarios’ platform is based on cutting-edge research in quantum computing and aims to empower both individuals and companies to leverage the potential of quantum computing. With Aqarios, using a quantum computer is no longer a daunting challenge reserved for a select few. 

Learn more about the future of quantum algorithms for solving optimization problems from our interview with the co-founder and CEO, Michael Lachner

Why Did You Start Aqarios?

The QAR-Lab of Ludwig Maximilian University of Munich is a research lab founded in 2016, from which we spun off in 2021 after seven years of research on quantum machine learning and leveraging quantum computing to solve optimization problems. The QAR Lab focuses mainly on fundamental research and has collaborated with about 20 German DAX companies. We also work closely with industry players at Aqarios. 

I come from a computer science background, but I have always had a keen interest in physics, especially astrophysics, general relativity, and, of course, quantum physics. Quantum computing combines both my interest in computer science and quantum physics. When this opportunity opened up, I joined Aqarios as the company’s first employee, and in mid-2022, I took on the role of CEO.

How Do You Build a Platform for Solving Optimization Problems?

We want to bring quantum computing to real-world applications, in particular, using it to solve optimization problems. Our platform can be accessed in different ways, including a graphical user interface, an API, and the option to upload optimization problems in an established industry format, for instance, if they were formulated for Gurobi.

Once you upload an optimization problem, it is stored and analyzed internally. We then suggest the best way to solve it, relying on a mixture of AI, domain knowledge, and heuristics to give the best recommendation. This takes into account a ton of different quantum algorithms and quantum hardware platforms.

Our platform speaks the language of quantum computing. We formulate the problem in a way a quantum computer can understand, without requiring the end users to have an in-depth understanding of the process. After the solver is chosen, we send the job to a quantum computer to calculate a solution. Once it comes back, we map the raw output of the quantum computer back to the original problem and present it in a format the end user can understand and act on. 

We develop solvers in-house, mainly inspired by research and client projects. We also incorporate open-source solvers for benchmarking purposes and run proof-of-concept projects to evaluate a solver’s performance. Our platform is designed in such a way that other developers can offer their third-party optimization solvers. The first collaborations have already started.

Since quantum computing is still very early, we also provide classical, quantum-inspired solvers and hybrid quantum-classical solvers that involve a mix of classical and quantum computing. Right now, there are still obstacles to overcome to use quantum computing in production, but it definitely will be used in the future, and we can already evaluate its future potential today. 

For now, it’s a lot of trial and error to see what works best. For practical applications, quantum annealers that perform analog quantum computations are more advanced in size than digital quantum computers that can run a gate-based model. When we tell them about today’s quantum computers, many companies looking for production-ready solutions are initially startled by the small problem sizes that can be solved. But large optimization problems are simply very hard, even for classical computers.

We’ve partnered, for example, with Gurobi, one of the leading and most efficient providers of classical solvers, to address this issue. Yet, beyond a certain problem size, even with high-performance computing hardware, it gets notoriously hard to find a good solution for large optimization problems. 

That’s why today, a lot of heuristics are used to find somewhat good solutions, but we see great potential in quantum computing to find solutions that you couldn’t compute classically in a reasonable time or simply faster than even today’s supercomputers. As quantum computers mature, they will become much more powerful and able to address even larger problems. It will be exciting to see how big the advantage will be, which many publications and empirical results already show to be quite promising.

How Did You Evaluate Your Startup Idea?

The idea for Aqarios came naturally, given our background at the QAR-Lab, with its cutting-edge research. By working closely with the industry, we learned what companies are interested in, what is already out there, and what we could bring to the table. We steadily evaluated this, and working with larger companies like SAP, BASF, EON, BMW, and Siemens helped us a lot. 

I will give you a specific example. We talked to a major company that wanted to find out what quantum computing could offer today. So we worked on their use case and provided different implementations and benchmarks on actual quantum hardware. For comparison, we also developed a classical implementation. While we initially wanted to focus on the quantum solutions, we found that even this classical implementation was better than what they used in production. This shows that there was a lot of room to optimize things—and lack of knowledge of how to do so—even in the classical realm. 

It’s often not enough to simply throw one algorithm at your optimization problem and stick with it—different use cases might require different approaches, and this may evolve over time. And that’s one of the things we want to enable users to do with our platform, by offering the best up-to-date optimization solutions from both the quantum and classical worlds.

What Advice Would You Give Fellow Deep Tech Founders?

Especially for deep tech founders, it’s important to keep two things in mind. One is the technology, for instance, an algorithm you’re building. The other one is the market and the customers you’re targeting and what they need. A market without a technology is useless just like a technology without a market. People coming from a tech background often forget that finding a good market is equally important, so go out there, identify the market, and talk with your customers, regularly. 

Comments are closed.