Software Campus grant for two BIFOLD PhD students

Lennart's project, "A System for Collaborative and Decentralized Federated Learning", aims at developing a system for end-to-end Federated Learning (FL) across organizational boundaries. Current FL systems do not solve the data management challenges of cross-silo FL, such as data discovery, data integration, or system-inherent privacy guarantees. Therefore, in order to fully leverage FL's capabilities for training machine learning models in a privacy-preserving manner, we need a new system design that supports organizations throughout the entire process of implementing FL use cases. In his project, Lennart will be working with Volkswagen, one of the largest car manufacturers in the world.

Haralampos's project, Efficient and Effective Query Processing for Federated Data Analytics, aims at assisting today’s data scientists in creating valuable insights by breaking inter- and intra-organizational data silos. Current mediator-wrapper-based solutions are based on a centralized query processing model and hence require additional resources, lead to redundant data transfer, and cause longer execution time. This project proposes a novel query processing architecture where participating database management systems collaboratively compute user query results in a decentralized fashion. For his project, Haralampos will be collaborating with Huawei, a large telecommunication company and cloud service provider.