PhD in data science focusing on privacy-preserving federated learning infrastructure – joint...

Updated: about 1 month ago
Job Type: Temporary
Deadline: 08 Nov 2020

The Clinical Data Science group at Maastricht University and Centraal Bureau voor de Statistiek / Statistics Netherlands (CBS) have a vacancy for a joint PhD student. At the Clinical Data Science group, we work on three main research areas: building global data sharing infrastructures; applying machine learning to build models from these data; and, using these models to improve healthcare. CBS is the Dutch national statistical office, and provides reliable statistical information and data to produce insight into social issues, thus supporting the public debate, policy development and decision-making while contributing to prosperity, well-being and democracy.

The successful candidate will dive into the thriving research field of federated learning, that studies how to train models from decentralised datasets. Federated learning has attracted the attention of major AI players like Google and Nvidia because it provides a workaround to the growing ethical and legal barriers to data sharing. Federated learning has the potential to play a key role in the next few years in the evolution of privacy-conscious, data-hungry AI, as shown by the fact that major AI conferences (e.g. Neurips, ICML) are already holding workshops dedicated to it.

This position is embedded in the CARRIER project, funded by the NWO’s Commit2Data round for Big Data & Health. The CARRIER project targets primary and secondary prevention of coronary artery disease (CAD) with a regional alliance of clinicians, citizens, legal experts, and data scientists. In CARRIER, patients will co-create a personalised health plan in collaboration with clinicians, adherence to which will be supported by an electronic lifestyle coach (eCoach). We will combine clinical data from different sources with socio-economic data, such as CBS, to build predictive and prognostic models that will detect people at risk of CAD and drive the behaviour of the eCoach.

In this project, the successful candidate will be responsible for developing the federated learning solutions for horizontally and vertically partitioned data, necessary to build the prediction models that will drive the intervention from different data sources. In doing so, they will face multiple challenges arising from the complexities inherent to privacy-preserving federated computing and the analysis of distributed heterogeneous data. The PhD project will lead to practical applications in the official statistical domain and the candidate will undertake research on how a national statistics office such as CBS can help improve healthcare.

This joint PhD position is also linked to the interfaculty institute BISS (Brightlands Institute for Smart Society) team consisting of a multidisciplinary team of researchers. One of the focuses of BISS is developing privacy-preserving techniques to connect distributed data and creating value by increasing the use of existing data.

View or Apply

Similar Positions