Jeannie Cahalane
Sylwia Olbrych is building a decentralized decision support system that helps companies collaborate on sustainability without exposing sensitive data. Her research at RWTH Aachen is turning complex circular economy goals into practical, peer-to-peer learning across the supply chain.
1. What inspired you to create a decision support system focused on the circular economy, and what problem were you most determined to solve?
Sylwia Olbrych: The current environmental situation requires us to rethink how products are designed, recycled and reused, and these decisions depend on so many interlinked variables that it’s almost impossible for us humans to handle them all at once. I view this challenge as a giant puzzle – one must break the problem down into smaller, manageable pieces and then bring them back together to see the full picture. But without the right tools, it’s easy to miss connections or make wrong choices. The improvements that are small can bring great benefits, and that should not be overlooked. My goal was to design a system that makes it easier for different companies to work together securely, learn collectively, and ultimately speed up innovation and implementation of circular practices.
2. How has your research at Laboratory for Machine Tools and Production Engineering (WZL) of RWTH Aachen University shaped your approach to building a system that balances collaboration with data privacy?
Sylwia Olbrych: While working at WZL of RWTH Aachen, I have been involved in multiple projects with various industrial applications, such as KIOptiPack and FAIRWork. I noticed that many companies and academic institutions face similar challenges and wish to collaborate, but they often lack the necessary technical tools. I also learned not only how such a system should be built but also how to integrate human and sustainability aspects into its design.
Our institute is involved in a wide range of projects and offers an interdisciplinary environment. That allowed me to gain insights from other groups which focus on sustainability, human-AI interaction, and process quality management. Some might think that all these diverse perspectives would slow down the research, but in fact, they gave me ideas I would not have discovered otherwise, ultimately strengthening my work.
3. What are the biggest technical challenges in enabling companies to share sustainability knowledge without exposing sensitive competitive data?
Sylwia Olbrych: In any collaboration, the key is to gain new insights and successfully integrate them into local processes without compromising sensitive information. The biggest technical challenge is finding the right trade-off between what type of data — whether it’s model predictions or learning gradients — a company is willing to share and the accuracy of the results. In the methods I’m investigating, it’s challenging to create a fast and secure way to communicate, including instant model fusion and real-time visualisation of the results. Handling heterogeneous data and models adds another layer of complexity. These are just a few examples of how complex building such a system can be.
4. What were some key moments in your journey to develop this technology, particularly in identifying the initial use cases in the battery and packaging sectors?
Sylwia Olbrych: At the beginning of my research, like many researchers, I needed time to find the direction for my research. I concentrated on the projects I was responsible for, and after a year, I could clearly see where my early ideas could be applied in practice. Identifying the battery and packaging sectors as initial use cases was a key turning point. Both sectors face significant global challenges related to reusability, recycling, and giving materials a second life. There’s an urgent need to design products and optimize processes that make it easier to recover valuable materials and keep them in circulation for as long as possible. The idea of contributing to the circular economy really resonated with me, and I was motivated by the fact that the solutions I was developing could help extend product lifecycles and reduce waste in a meaningful way.
5. Can you walk us through how your system works? How does it use collaborative learning to support sustainability-driven decisions across different companies?
Sylwia Olbrych: Let’s imagine that you are a process engineer at a compounding extrusion company and you would like to incorporate more recycled polymer material into the process while maintaining the product quality required by end users and EU regulations. You can perform additional tests, or you can collaborate with another compounding extrusion company to reach the goal faster. This is exactly where the system could support companies. Using the system, both companies independently train local ML models using their process data and exchange model gradients iteratively. The system should enable them to follow the modelling and access the actionable recommendations to improve their manufacturing processes.
6. What sets your decision support system apart from other sustainability or supply chain management platforms?
Sylwia Olbrych: Unlike traditional sustainability or SCM platforms, which often require sensitive operational data to be centralized for analysis, the decision support system design focuses on peer-to-peer collaborative learning, avoiding a single point of failure by eliminating a single controlling entity. This is made possible through a decentralised approach that does not rely on a centralized server for global model training. Such an approach also offers flexibility, enabling various cross-company or cross-sector collaborations, which are essential for true circular supply chains.
7. How will your tool help Small and Medium-sized Enterprises (SMEs) navigate complex EU regulations on waste, design, and reuse?
Sylwia Olbrych: I’ll answer this question using an example from the packaging sector. The regulations introduced by the EU significantly affect the required amount of recycled material in plastic packaging, with targets set to increase steadily over the coming years. By 2030, packaging will need to contain between 10% and 35% recycled content, depending on the type. For example, PET bottles must have at least 30% recycled material. By 2040, that number is expected to rise to as high as 65% for single-use plastics such as beverage bottles. These regulations are pushing companies to develop new products and processes quickly, which can be challenging for SMEs to do alone. Such a system can help SMEs by connecting them with trusted partners like universities and other companies to develop new solutions.
8. Can you share a success story or a specific example that illustrates how your system could help two companies collaborate to optimize a manufacturing or recycling process?
Sylwia Olbrych: We have recently conducted a study focused on improving prediction accuracy for the state of health of Li-ion batteries, using electrochemical impedance spectroscopy data. The results show that our collaborative approach enables knowledge exchange between two CNN-BiLSTM models, which leads to better-performing models after the models’ convergence. This is really promising because it means we can achieve more accurate estimates of remaining useful life for batteries, which directly supports better planning for second-life applications, which is a crucial aspect of sustainable battery reuse and e-waste management. I am also happy to share that the results of this use case are planned for publication.
9. What challenges have you faced while developing this system, and how did you overcome them?
Sylwia Olbrych: There were two main challenges that I faced. The first challenge came up during the investigation of a use case involving the prediction of battery state of health, which required working with two distinct datasets. We realised that the measurements were carried out on two different battery types and at temperatures that did not always overlap. Instead of stopping the investigation, we applied a collaborative learning approach. It was encouraging to see the improvement in prediction for both models.
The second challenge emerged when the entire process chain needed to be combined to enable process optimization with the goal of increasing the share of recycled polymer in the final product. This was complex because each process step was managed by a different company. Developing a clear concept for collaboration across partners was difficult at first, but this experience showed me the importance of designing a flexible approach.
10. Looking back, are there any strategic decisions in your research that you would approach differently now?
Sylwia Olbrych: If I could go back, I would have involved end users in co-design workshops to incorporate their insights into the first system design phase. I believe this could have helped me design it faster and in a more user-centric way from the start. The research is still ongoing, and I plan to include end users in the design evaluation so I can see what could be improved or removed from the system.
Finally, I would expand the system’s architecture to handle multi-party collaboration. Right now, the design mainly supports peer-to-peer connections and has been tested on two models. However, the ability to involve multiple companies at once would be beneficial for complex supply chains.
11. What does success look like for you and this technology in the next 5–10 years? What is your ultimate vision for an intelligent, collaborative sustainability network?
Sylwia Olbrych: For me, success over the next five to ten years means establishing an intelligent and collaborative sustainability network that effectively connects diverse partners within industries. I imagine a user-friendly online platform where companies and academic organisations can easily identify and engage with trustworthy and suitable collaborators, based on clearly defined sectors and use cases. Afterwards, users should be able to find the selected partners in the decision support system installed locally on their machine and begin collaboration at the model level.
Such a network could change the way we approach collaboration by offering solutions that link real-world scenarios, collaboration and technology.
12. What does a typical day in your life as researchers look like, and how do you stay motivated?
Sylwia Olbrych: As a researcher at WZL of RWTH Aachen, I participate in many projects where I take on various roles and responsibilities. My days are filled with discussions and workshops, making my work anything but dull! I also engage in teaching activities, such as supervising master’s and bachelor’s theses and guiding group projects. I genuinely enjoy knowing that my feedback can motivate students and assist them in overcoming obstacles when they feel stuck. These moments continue to motivate me to advance with my doctoral project.
Editor’s Note
Olbrych’s work bridges AI, data privacy, and real-world manufacturing. Her system supports companies facing growing EU regulations by enabling secure knowledge sharing and model-level collaboration. This interview explores how technical design and policy awareness come together to accelerate circular practices.
