Limassol startup develops platform for evaluating quantum computing — Cyprus Mail

Cypriot startup Superpositions Studio, operating in Limassol, is developing a software platform to assess whether quantum computing can be useful for specific business tasks. As Cyprus Mail reports, the service allows users to describe a problem in plain language, after which the system selects quantum and classical algorithms, conducts experiments and compares the results.

The company says that for standard tasks, the platform can prepare a conclusion in less than an hour. The client receives a report on which approach proved more effective in terms of accuracy, speed and computing cost, as well as Python code for rerunning and adapting the solution. A private operating mode is available for companies with sensitive data, under which information does not leave their environment.

Comparison with classical models

Superpositions Studio emphasizes that every quantum experiment is accompanied by a classical baseline model. This is intended to allow clients to see whether a quantum or hybrid approach offers an advantage specifically for their task. The company noted that sometimes the most useful result may be the conclusion that a problem is not yet suitable for quantum computing.

The platform uses a multi-agent artificial intelligence system with dozens of specialized agents. It analyzes the task, transforms it into a mathematical model, writes code and performs calculations on high-performance simulators or, when necessary, on quantum processors.

More current news is available on the UA.News Telegram channel Telegram.

Tests in finance and energy

According to the company, the platform has been tested in more than 20 industry scenarios and with 10 quantum algorithms. Areas include payment fraud detection, transaction screening for anti-money laundering, risk assessment and wind energy generation forecasting.

In one wind power generation forecasting test, a hybrid quantum neural network was trained on an industrial dataset with more than 26,000 observations. The company said the model showed a forecast error of 7.1%, compared with 7.3% for a classical neural network, while using approximately one-third fewer parameters.

Superpositions Studio also notes that its service is not tied to a single quantum hardware manufacturer: experiments can be run on systems by IBM, IonQ, IQM and Rigetti. The company said it has more than 150 users in 20 countries.