Together with experienced professionals in our STRAD MD project team, you will work with the STRAD MD federated learning platform. This platform is being set up at the UMC Utrecht and at one or more locations of the large mental health organizations Arkin and Parnassia Group so that research projects involving sensitive data from large and diverse groups of patients can safely be utilized for machine learning projects. The goal of this project is to model dynamic interactions between symptoms of depression, patient properties and interventions (such as medication choices and therapy sessions), with the data involving a lot of free text.
The STRAD MD project will form the basis for a more sustainable way of collaborating with data, preferably in a federated manner, between mental health institutions in the Netherlands. It is your challenge to help realize this growth, especially regarding developing and implementing durable NLP solutions. In our team you will have a lot of say in designing and implementing appropriate solutions, plenty of room to experiment, and every opportunity to develop yourself.
Your activities at a glance:
- Developing and setting up (federated) NLP pipelines across three mental health institutes for feature/ entity extraction on the federated learning platform (other project members will work on technically setting up the federated learning platform, devops/ database engineering is not part of the postdoc). You can build on work of our previous NLP PhDs and postdocs, collaborate with the UU teams involved and/ or bring your own flavor.
- Modeling dynamic interactions between patient properties, symptoms, intervention preferably using explainable approaches, partly based on the features/ entities extracted with the NLP pipeline.
- Exploring the combination of combining the two techniques above in one optimization pipeline.
- Exploring the possible added role of synthetic data for facilitating federated learning.