PhD - Efficient Neural Representation of Datasets

Do you want to turn your ideas into useful and meaningful technologies? Whether in the area of Mobility Solutions, Consumer Goods, Industrial Technology or Energy and Building Technology - with us you improve the quality of life of people around the world. Welcome to Bosch. Robert Bosch GmbH is looking forward to your application!

Type of employment: Limited Working time: Full time Place of work: Renningen We are researching state-of-the-art Deep Generative Models that are used to make real Bosch systems data-efficient. We are looking for a:n doctoral student who is interested in researching creative applications of generic models (e.g. stable diffusion) as a controllable data set representation for training and validation of networks for downstream tasks. Not all data points in a dataset are equally important for the performance of a neural network. As training progresses, the loss of some data points may no longer be meaningful, as the network has already learned everything it can. Therefore, it may be advantageous to observe network training to provide the right data at the right time. However, the selection of data from a fixed dataset is problematic if there is no image with the exact mixing of attributes. The aim of this doctoral project is to develop new learning algorithms for generating relevant data on demand in response to the network's needs. This includes, among other things, improving training efficiency through the synthesis of data of better relevance, as well as ensuring that the desired invariance is achieved by specific examples. As part of our team, you will develop new approaches to adapting deep generative models (e.g. diffusion models, GANs, UAEs) as data sources to better train and validate downstream models. In addition, you use the controlability and knowledge of generic basic models to no longer consider records as a loose collection of images. They discuss and develop new ideas with the experts Experts for deep learning and computer vision at the Bosch Center for AI. In addition, you have the opportunity to publish your results in high-level journals and conferences.

Education: excellent completion in computer science or a related area focusing on computer vision and deep learning Experiences and know-how: sound background in deep learning and computer vision, experience with deep learning frameworks (TensorFlow, PyTorch, etc.) and very good programming skills (especially Python), practical experience with Deep Generative Modeling and Foundation Models are a plus, publications of peer-reviewed research work also of advantage, good scientific writing skills Enthusiasm: Enthusiasm to work in an interdisciplinary and international team Languages: very good English

Work-Life Balance: Flexible work in terms of time, location and work model. Health and Sports: Wide range of health and sports activities. Childcare: childcare services. Employee discounts: Advantages for employees. Free space for creative work: free space for creative work. Social counselling and care: Social counselling and brokerage services.

The future personnel department or the department will be happy to inform you about the individual service catalogue. JOBV1_EN