Postdoctoral Researcher AI4Hearing

About us

The School VI of Medicine and Health Sciences comprises the fields of human medicine, medical physics and acoustics, neurosciences, psychology and health services research. Together with the four regional hospitals, School VI forms the University Medicine Oldenburg. Furthermore, the university cooperates closely with the University Medicine of the University of Groningen.

In the AI4Health division, methodological questions in the areas of self-supervised/label-efficient learning and explainability of deep neural networks (XAI) are being developed, particularly for use in biomedical applications. Further information about the division can be found at https://uole.de/en/ai4health

The position is part of the DFG-funded cluster of excellence "Hearing4All.connects". Among its research goals is uncovering the multifactorial patterns of hearing disorders through a novel combination of knowledge- and data-driven approaches.

Your tasks

One goal of the research area "data-driven personalization of hearing support" the position resides in is to build up technology for data-driven pattern recognition to infer data-driven generalized auditory profiles. The central challenge for this endeavor are the small sample sizes (of typically a few hundred samples or less) of the tabular datasets under consideration. This challenge is supposed to be targeted through the use of tabular foundation models. The following specific tasks are supposed to addressed:

• Characterization of the inductive biases inherent to tabular foundation models
• Adaptation of tabular foundation models through continual pretraining (on synthetic but also on real data) to the data distributions at hand and evaluation of their predictive performance
• Comparision to alternative approaches applicable in the small-sample-size regime such as few-shot learning, meta-learning, transfer learning or probilistic methods

This would evaluate and potentially enhance the usability of deep learning models for complex tabular data with mixed continuous and categorical columns with non-Gaussian feature distributions as a stepping stone towards the development of generalized auditory profiles. Tabular foundation models may also help to improve the derivation of Common Audiological Functional Parameters (CAFPAs), which have been proposed as a means of integrating different databases containing various audiological tests and as an intermediate layer of a clinical decision support system for audiology. 

Your profile

Required qualifications:

• PhD or equivalent achievement (proof of independent research capability) in Machine Learning, Computer Science, Physics, Mathematics, or a related field
• Deep theoretical knowledge and extensive practical experience in machine learning, evidenced by publications, personal projects, or professional work
• Expert-level proficiency in Python and PyTorch
• Strong publication record in peer-reviewed venues (conferences or journals), demonstrating independent research capabilities
• Excellent English language skills, both written and spoken

Preferred Qualifications:

• Experience working with tabular data and structured datasets
• Hands-on experience with foundation modelsExpertise in small-sample learning methods (e.g., few-shot learning, meta-learning, transfer learning)
• Proficiency in German language

We offer

• Payment in accordance with collective bargaining law (special annual payment, public service pension scheme, asset-related benefits) incl. 30 days annual leave
• Support and guidance during your onboarding phase
• A family-friendly environment with flexible working hours (flexitime) and the possibility of pro-rata mobile work
• Benefits from the university's health promotion programme
• An extensive and free further education programme as well as programmes geared toward the promotion of early career researchers (https://uole.de/en/school6/early-career)

Our standards

The University of Oldenburg is dedicated to increase the percentage of female employees in the field of science. Therefore, female candidates are strongly encouraged to apply. In accordance to § 21 Section 3 NHG, female candidates with equal qualifications will be preferentially considered. Applicants with disabilities will be given preference in case of equal qualification.

Further information

The division AIHealth is headed by Prof. Dr. Nils Strodthoff. 

Opportunities for further academic qualification (Habilitation) are available.

Contact

For further information, please contact Prof. Dr. Nils Strodthoff ([email protected]).