My passion lies in analysing and designing intelligent systems that go beyond performance, emphasizing properties of trustworthiness, particularly privacy and explainability. I then like to apply these to medical use cases in the modalities of images, texts, and graphs. In my PhD, I will focus on designing and improving lung cancer diagnosis and future risk prediction, as well as on privacy-preserving and trustworthy machine learning.

I like to work in two overlapping realms:

• Understanding the privacy vulnerabilities of a model.
• Enhancing the privacy of individuals whose data is used to train the models.

In the first realm, I am interested in the memorization of data by ML models, and reconstruction attacks. My interests in the second realm include applying differential privacy under realistic threat models, machine unlearning, and exploring privacy with explainability.

I am open to collaborations in these areas. While I currently do not have any open topics for IDP/Application project, Guided Research, or Thesis, I encourage motivated students interested in these areas to reach out with their specific interest and motivation.