Academic Assistant - Deep Learning & Embedded AI in the research project Lab2Device

Self-learning of research tasks in the Lab2Device project Independent publication of research results from the Lab2Device project Student guidance (e.g. Hiwis) Self-transfer of research results in working applications Modern deep learning models are powerful but resource-intensive. This prevents their local use on embedded systems and leads to inefficient, unsafe and data protection-critical cloud-based reference approaches. Model Compression (e.g. quantization, Pruning and Knowledge Distillation) and Neural Architecture Search (NAS) approaches allow the local use of deep learning models, but often lead to a reduction in the model performance and are associated with several other limitations (e.g., high computing complexity and thus high energy costs for neural architecture search approaches). The aim of this doctoral thesis is to research new Model Compression techniques that reduce the relevant costs for use on embedded systems (such as storage consumption, latency and energy consumption) with minimal loss of model performance. You work in an interdisciplinary team (three research institutes of the University of Offenburg are involved: IMLA, INES, ivESK) with another doctoral student (focus on neural architecture search) and two postdocs to enable and benchmark resource-efficient deep learning models for local inference on embedded hardware.

Completed scientific higher education (master or comparable) in the field of computer science, statistics, electrical engineering or information technology Experience: Detectable programming skills (Python, C/C++) Very good knowledge and practical experience with modern machine learning methods and frameworks (PyTorch and/or JAX) At least basic knowledge of embedded systems Basic German knowledge of advantage, good English knowledge meaningful

Personality: self-initiative and structured working Interest in creative problem solving Joy of interdisciplinary teamwork

attractive training courses that you can continue in your job a modernly equipped workplace Special payment flexible working hours Corporate Benefits Access the possibility of children's holiday care in the academy Kita Sommersprosse attractive offers in the context of occupational health management and the reconciliation of work, family and care