Future-Proof Decision Intelligence With Quantagonia’s Hybrid Quantum Platform

Enterprises of all sizes face complex decision-making challenges every day, from optimizing capital investments and streamlining staffing schedules to enhancing production efficiency, revolutionizing logistics, and improving overall performance.

Finding the single best solution is often impractical due to the vast array of potential options. Therefore, the objective shifts to identifying an optimal solution given the constraints of time and computing power.

Quantagonia has introduced the world’s first hybrid quantum platform to enhance enterprise’s decision-making, combining advanced classical solvers with future-proof quantum computing capabilities. Since our first interview in spring 2023, the company has secured a seed round of €4.3 million, led by Tensor Ventures and followed by existing investors Voima Ventures, FTTF, the Beteiligungs-Managementgesellschaft Hessen, and a German family office.

We had the pleasure of speaking again with the founders, Dirk Zechiel and Philipp Hannemann, about the ins and outs of solving optimization problems, whether improving the computing hardware or the algorithms yields better improvements, and who is using their hybrid quantum platform* already today. 

What Makes Your Hybrid Quantum Platform Special?

Since the last time we spoke, we have developed our first product: It’s a fully cloud-based platform where users can submit their optimization problems in standard formats through a graphical user interface, and the platform automatically chooses the most suitable backend. It is hybrid as it uses classical processors like CPUs and GPUs and will leverage quantum processing units (QPUs) in the future.

QPUs have yet to be available at scale. However, we have a proof-of-concept using IBM’s quantum machines, demonstrating that we can use them to solve parts of an optimization problem faster and better. While it’s not commercially viable yet, we can estimate as quantum computers get more powerful what they will be able to do in the future and what customers can expect. 

Enterprises need solutions for their optimization problems today; they can’t wait for quantum advantage. But in the not-too-distant future, we’ll reach quantum advantage, and we should prepare for it. It’s like working on AI before it was cool, five years before ChatGPT came out.  Now, everyone gets the potential of AI and needs to do something about it—and the same will happen with quantum computing. With our hybrid quantum platform, we can solve optimization problems today and harness quantum computing over time.

Importantly, we’re the only quantum software provider that can solve generic optimization problems. Most solvers use heuristics that work fairly well for specific problems and constraints, but for real-life applications, these change all the time: new constraints come up, or the number of input variables changes. With our generic approach, we don’t need to find new heuristics every time. Instead, we abstract away the formulation of an optimization problem and thus can tackle new ones at the press of a button.

We can solve optimization problems of any size—it’s just a matter of how much compute time we can spend and how satisfied a user is with a found solution at any time step. Generally, problems involving hundreds of thousands to a few million variables are not a problem at all (pun intended!).

Also, the size of an optimization problem doesn’t correspond to its complexity. Some problems may have many variables but aren’t complex and are thus easier to solve than the ones with fewer variables but a high degree of interdependence between those and thus a larger degree of complexity. Our platform analyzes each optimization problem, removes redundancies, identifies sub-problems, and whether heuristics apply, making it smaller and easier to solve.

How Do You Evaluate and Improve Different Solvers?

There are publicly available benchmarks for predefined optimization problems, which track the time required to solve them based on different methods and machines. Our classical solvers already match the performance of leading commercial solvers like CPLEX and Gurobi. But in addition, our quantum solvers also ensure we are future-proof.

Since we can choose between different backends, we can utilize all of them where they work best: GPUs excel at performing simple operations rapidly and in parallel, but they can’t handle if-then scenarios and split decisions. These can be managed by other processors, or in some cases, it’s even more efficient to avoid if-then decisions altogether and execute both scenarios simultaneously, given the incredible speed of GPUs.

Currently, the greatest improvements in solving optimization problems come from the software side. Given the complexity of these problems and the exponential number of potential solutions, investing time and effort in enhancing algorithms and decomposing the problem yields better results than simply upgrading hardware.

Computing hardware improves every year thanks to Moore’s law. But if it takes five years to solve an optimization problem, even if microprocessors improve twofold or if you buy a second computer, it would still take years and thus far too long to obtain a solution. That’s why most hardware improvements won’t move the needle for compute problems with exponentially many possible solutions. 

The only exception is quantum computers, which have the potential to solve some computational problems exponentially faster. This will change the game entirely because quantum computers will speed up computations not just by a constant factor like 2x and can thus play in an entirely different league compared to classical computers.

Who Will Benefit From Your Hybrid Quantum Platform?

We target three different customers with the platform:

One is consulting companies like McKinsey, Accenture, or KPMG, which help clients solve important decision problems, whether it’s which aircraft needs to fly which route or which factory to build. We help them formulate and solve these optimization problems more straightforwardly, thus helping their clients make better decisions. 

The next one is helping end users working at large enterprises like Lufthansa or Deutsche Telekom to make complex decisions as part of their operations research department. They have the in-house expertise to model these optimization problems, and we can provide them with better and faster solvers to do the job. 

Finally, independent software vendors like Oracle, SAP, or Salesforce will include our platform in their software offerings, e.g., as a decision-making module bundled with an ERP system. 

Solving optimization problems requires expert knowledge and human experts to formulate these problems properly, and that’s our current focus with the platform. But, in future releases, users will be able to describe their optimization problem in natural language, e.g., to add another shift to the schedule. Then, large language models will formulate it properly so our solvers can tackle it. This will open up new markets as people who aren’t familiar with operations research will be able to formulate and solve these problems.

I have been working in optimization for the past 25 years, and the need is still very clear. Often, people ask why optimization and decision intelligence aren’t used everywhere. The answer is that the need for human experts creates a massive hurdle. AI won’t be a good fit for solving optimization problems—creating new schedules will only be as good as your training data—but AI will be a game changer for formulating optimization problems and then handing them to the right solver. 

Let me give you an analogy: Currently, solvers are like the engine in a race car that only Formula-1 drivers can drive. We’ll take the engine out, place it in every car, and use AI to make the car self-driving so everyone can use it. Empowering employees at large enterprises to solve tough optimization problems will be our core focus. 

What Was One of Your Key Learnings Since We Last Spoke?

Work on a clear roadmap to get from an innovation to a product, and then put a strong team and execution in place to follow this roadmap. Building a deep tech startup is not just about technology but also about product management and vision. That’s why we have a team of four people dedicated to guiding our product development. And we use an agile development process, two-week sprints, and a mix of OKRs and Scrum to drive our product development roadmap forward. 

What to Learn More?

Check out Quantagonia’s Hybrid Quantum Platform* to register with just a few clicks and solve optimization models free of charge in the cloud with their HybridSolver.


*Sponsored link—we greatly appreciate the support by Quantagonia

Comments are closed.