AI/ML 
Rozwój oprogramowania 
Artificial intelligence is often presented as the answer to nearly every data analysis challenge. In practice, however, effective solutions do not always rely on AI models. In some cases, carefully selected mathematical algorithms deliver far greater value because they are faster, simpler, and better suited to the specific application.
This is one of the key takeaways from the latest episode of the “AI in Production” podcast, featuring Rafał Pisz from Quantup. During the conversation, he discusses the Peak Detection project completed for a manufacturer of specialized laboratory equipment. The story shows that innovation does not always mean using the newest technology. Often, the greater challenge is understanding users’ actual needs and designing a solution that makes their everyday work easier.
When the problem is not the technology, but how it is used
The laboratory equipment manufacturer involved in the project had been supplying systems for advanced chemical analysis for many years. The device itself worked as expected. However, as the company began planning expansion into new markets, an important question emerged: how could the product also be used by people with less experience in interpreting measurement results?
The challenge was not to modify the device’s parameters or increase its computing power. The real barrier was the need to analyze the charts generated by the system manually. For experienced specialists, this was a routine task. For new users, however, it required specialist knowledge and practical experience. This stage was limiting the product’s broader adoption.
As Rafał Pisz explains, technological development often involves lowering the barrier to entry. A device should not require users to develop increasingly advanced skills. On the contrary, it should take over some of their tasks and make the process easier.
Peak Detection: automating result interpretation
The project focused on automatically detecting characteristic peaks in measurement charts. Correctly determining where a signal began and ended was essential for the subsequent interpretation of the test results.
Although this may appear to be a straightforward task, this type of data analysis can be highly demanding in practice. Even experienced operators may disagree on the exact boundaries of a signal, and a small shift can affect the final result. Automating the process therefore did more than accelerate the work. Above all, it ensured repeatable results and reduced the dependence on the experience of the person operating the device.
Why artificial intelligence was not used in this project
One of the most interesting parts of the conversation is the explanation of why the Quantup team deliberately decided not to use artificial intelligence models.
Instead of machine learning, the team applied carefully selected deterministic algorithms and mathematical methods that were better suited to the nature of the problem. This approach delivered high effectiveness while keeping hardware requirements low and ensuring fully predictable operation.
This is an important lesson for organizations planning investments in AI. The starting point should always be a clear understanding of the business problem and the expected outcome. Only then should the technology be selected.
In some projects, artificial intelligence will be the right choice. In others, conventional mathematical algorithms will perform better. The success of a solution depends on its effectiveness, not on the popularity of the technology behind it.
Hardware limitations require sound design decisions
Another challenge was the environment in which the solution had to operate. The algorithms were not running in a cloud environment with virtually unlimited resources. They had to run on a Raspberry Pi-class computer integrated into the laboratory device.
This limitation required a highly deliberate approach to software design. Every library, operation, and section of code affected memory consumption and overall system performance. As a result, the designers had to select solutions that were not only effective but also exceptionally lightweight and optimized for the available resources.
Technology as a tool for business growth
The most important outcome of the project was not the automation of data interpretation itself. The implemented solution allowed the manufacturer to offer its devices to a much broader audience because operating them no longer required the same level of experience.
This opened the door to expansion into new markets and increased the product’s sales potential.
It is a strong example of how the value of Data Science and automation projects extends beyond the technical dimension. Well-designed solutions can directly support product development, improve competitiveness, and contribute to business objectives.
Listen to the full conversation
The Peak Detection project is only one of the topics covered in the episode. Kuba Orczyk and Rafał Pisz also discuss how to make informed technology choices, why not every challenge requires AI, and how the laboratory equipment market is evolving as automation progresses.
For anyone interested in practical applications of mathematics, Data Science, and artificial intelligence in industrial projects, the full episode offers valuable insight. The conversation is based on experience from a real implementation and explores not only the technical side of the project but also the business decisions that contributed to its success. If you’re interested in the details of the implementation, read the full case study on our website.
Design, Development, DevOps czy Cloud – jakiego zespołu potrzebujesz, aby przyspieszyć pracę nad swoimi projektami? Porozmawiaj o swoich potrzebach z naszymi specjalistami.
Jakub Orczyk
Członek zarządu / Dyrektor sprzedaży
Zamów bezpłatną konsultację