At the Institute for the Protection of Maritime Infrastructures in Bremerhaven, we research and develop innovative solutions to strengthen the resilience of maritime infrastructures and make them adaptable, safe and sustainable. In close cooperation with partners from research, industry and other maritime security stakeholders, we combine technological innovation with practical expertise and offer you the opportunity to work on pioneering projects.
What you can expect
As a researcher in the Situational Awareness and Cyber Security group within the Maritime Security Technologies department, you will research and develop innovative AI-based approaches to data processing in the underwater domain, with a focus on the transferability of machine learning models between different hydroacoustic sensors (e.g. side-scan sonar, multibeam echo sounder, side-scan sonar).
You will plan and carry out research projects to develop and evaluate transfer learning strategies, with the aim of increasing the efficiency and robustness of AI models whilst taking into account the heterogeneity of sensor data. This includes the design of shared feature backbones, the application of domain adaptation and multi-task learning, as well as the pre-processing of raw data using GANs or specialised filters.
You will conduct experimental investigations, analyse results using appropriate metrics and visualisations, derive new research questions from the findings and test your methods in an experimental maritime situational awareness system.
You will thus actively contribute to the development of flexible, adaptable and secure safety solutions for maritime infrastructure – with a direct link between basic research and practical application.
Your tasks
- Designing and carrying out research on transfer learning strategies for AI models in the underwater domain, particularly regarding transferability between sensors based on different physical principles
- Development and implementation of algorithms and models for the processing of sonar image sequences and time-frequency representations
- Application and evaluation of methods such as feature extraction, fine-tuning, domain adaptation and multi-task learning on real and synthetic datasets
- Carrying out pre-processing steps (e.g. beamforming, calibration, denoising, normalisation) and employing data translation techniques (e.g. GANs)
- Development of shared-feature architectures with common backbones and sensor-specific head modules
- Evaluation and visualisation of results (e.g. heatmaps, confusion matrices, domain shift metrics) as well as formulation of scientific hypotheses
- Publication of results at international conferences
What you bring to the role
- A completed scientific degree (Master’s / University Diploma) in Computer Science, Electrical Engineering, Mathematics, Physics or another relevant field
- In-depth knowledge of deep learning (particularly CNNs, Transformers, autoencoders) and its application to spatially and temporally structured data
- Practical experience with transfer learning, domain adaptation and evaluation in complex domains
- Programming skills in Python and C, as well as experience with PyTorch or TensorFlow
- Knowledge of data fusion, multi-sensor integration and cross-modal architectures
- Experience with Docker, CI/CD, computer networks and Linux systems
- Good English skills and the ability to clearly communicate complex research findings
We offer
DLR stands for diversity, appreciation and equality for all people. We promote independent work and the individual development of our employees both personally and professionally. To this end, we offer numerous training and development opportunities. Equal opportunities are of particular importance to us, which is why we want to increase the proportion of women in science and management in particular. Applicants with severe disabilities will be given preference if they are qualified.
We look forward to getting to know you!
If you have any questions about this position (Vacancy-ID 4653) please contact:
Jannis Stoppe
Tel.: +49 471 924199 43
or Aljoscha Windhorst
Tel.: +49 471 924199 43
