Quantagonia: Shaping the Future of Quantum Optimization

Whether it’s finding the shortest path for a courier delivering parcels or a new flight schedule for an airline after a storm, in both cases, there is limited time to find a solution.

These optimization problems occur in real-world situations where there are so many possible solutions that it’s practically impossible to try them all. Thus, instead of searching for the best solution, the goal is to find an optimal solution. Optimization problems are so complicated that even supercomputers struggle to find truly optimal solutions.

Quantagonia was founded in 2021 by Dirk Zechiel, Sebastian Pokutta, Sabina Jeschke, and Philipp Hannemann to harness the powers of quantum computers to find better solutions for optimization problems, and thus help to make the world run more optimally. It raised a pre-seed round in early 2022 from Voima Ventures, FTTF, and other investors. 

Learn more about the future of quantum optimization from our interview with the co-founder and CEO, Dirk Zechiel: 

Why Did You Start Quantagonia?

For the past 25 years, I’ve dealt with large-scale optimization problems, such as determining a railway company’s schedule or assigning employees to flights for an airline. Due to their large scale, these problems are impossible to solve by hand: it’s even hard for computers to solve them within a reasonable amount of time, let alone in real-time.

Many applications, however, require real-time solutions. For example, imagine a big storm hitting Germany, and all the trains need to be rescheduled within a few minutes—not hours or days—to avoid delays as much as possible. 

In the last twenty years, people have joked that if quantum computers were around, they might be able to solve these optimization problems, but to date, no quantum computer has outperformed a classical supercomputer in practical tasks. This might change sooner than expected. 

When I sold my previous company, Gurobi, in 2017, it felt like the next right step was to explore quantum computing. It took me a while to wrap my head around it, as developing for quantum computers is totally different from classical software development. There are no for-loops, and as of today, you need to build a physical model for the quantum chip. Performing a quantum computation depends very much on the specifics of the hardware. 

But at some point in the future, quantum computers that outperform any classical computer will be available. So then, what do you do with all the codes written for classical high-performance computing? Thinking about forward compatibility, someone has to develop a program that would make existing codes compatible with new quantum processors and the dramatic speedups they may achieve. 

So, I decided to found Quantagonia in 2021 together with Sebastian Pokutta, whom I worked with at ILOG, Sabina Jeschke, whom I knew from her time at Deutsche Bahn, and Philipp Hannemann, whom I worked with in a professional sports scheduling company.

How Does Quantum Optimization Work?

Let’s start by understanding what an optimization problem is. 

One of the most famous examples is the traveling salesman problem. Imagine you wanted to visit several locations in Berlin using the shortest path to all of them without visiting any twice. It’s a really hard problem, as the number of possible routes grows factorially with each new location. If you want to visit many, many locations, it becomes practically impossible to find the best solution. Instead, given finite computing resources, you’ll find a more or less optimal solution. 

Such optimization problems, where many solutions differ in how optimal they are—but none is really perfect—are commonplace in real life: just think of delivery services that need to deliver parcels within a given time window. 

Instead of attempting to solve the exact problem, which would be impractical, one tries to develop heuristics that achieve a fairly optimal solution within a short amount of time. For the traveling salesman problem, it would be, for example, to replace crossing lines while going from one destination to another. 

Quantum computing is then the next logical step to tackle these problems, as it offers a whole new paradigm. Working with quantum bits, called qubits, the fundamental processing units that can be “0” or “1” at the same time, quantum computers can exploit unique properties like parallelism and entanglement to gain computational advantages that classical computers simply cannot access.

Our goal with Quantagonia is to hide all the complexity behind quantum algorithms and obtain a speedup for classical optimization problems. Our platform will transform these problems so that a quantum computer can calculate an optimal solution that is much better than what classical supercomputers could achieve. 

This involves modeling the optimization problem in a domain-specific language, such as a QUBO or mixed integer program (MIP), and then following a hybrid approach. Pre-processing might still happen on classical CPUs, the heavy lifting of finding an optimal solution might be achieved by a quantum computer, and post-processing may be done on a traditional CPU. Orchestration, that is, selecting what type of hardware is most suited for running a specific part of the algorithm, may also be done on a traditional CPU. 

Today, different quantum computing hardware platforms are available, from superconducting and ion traps to photonic quantum computers. We’ll make all these abstract since users won’t care about the specific hardware but about better optimization results. We’re actively experimenting with the different types of quantum hardware and have partnered with many quantum hardware vendors to access their hardware for running use cases. 

How Did You Evaluate Your Startup Idea?

After working on optimization for more than 20 years, the team and I had quite some experience addressing this huge market. In practice, this meant talking to lots of prospects, receiving LOIs, and demonstrating speedups for solving optimization problems.

Today, businesses often abstract from their actual problems, solving smaller toy problems because they don’t have the computing power to solve the whole problem. But they’d love to model more realistic problem sizes, which is where our hybrid quantum-classical approach comes in.

Beyond optimization, our approach will also allow us to speed up machine learning and simulation in the future. 

What Advice Would You Give Fellow Deep Tech Founders?

My background is technical, as I studied computer science, but I quickly realized that I also have to focus on the business side. To build great products, market them, and still listen to your customers—don’t just focus on elegant algorithms; consider the value your customers receive. 

I’ve also learned a lot about talking to investors. The world has changed, and investors nowadays look for more traction—and proving traction for a deep tech startup is tricky. It may take several months to develop the technology in the first place, but it is important to find ways to show revenue early on.