Doctoral School for Dependable Electronic-Based Systems.
Electronic-based systems (EBS) make our world “smart” by combining advanced electronics and software, often in networked systems that interact with the physical world through sensors and actuators. Most EBS applications in production or transportation are safety-critical: EBS failures may cost human lives. We thus aim to find fundamental concepts and methods, but also application-oriented tools to make EBS dependable, where dependability summarizes attributes of a system allowing humans to trust EBS.
The PhD Project.
AI systems are increasingly used to support human-decision making on the edge. Resource constraints and the need to autonomously adapt to the surrounding contexts present challenges to operating embedded AI systems in a trustworthy manner and hinder their adoption in safety-critical control applications. This project will develop novel methods for improving robustness and optimizing machine learning models, integral part of EBS, under resource constraints, as well as designing testing methodologies for on-device machine learning methods. The value of the developed methods will be explored in several use-cases. One example is to continuously adapt the parameters of power amplifiers in the field to enable communications technologies providing faster transmission rates, higher-bandwidth and better signal quality.
Master’s degree in Computer Science, Computer Engineering, Information Technology or similar.
Very good knowledge of machine learning, deep learning (PyTorch, TensorFlow), understanding of computer architecture, compilers and embedded systems. Practical experience in any of these fields is a plus.
Interest in theory and practice of optimizing deep learning for resource-constrained devices. Interest in conducting top-quality research in this field. Prior research experience is a plus.
Solid programming skills in Python, C/C++.
Open-minded personality, eager learner, team player.
Our Research Group.
We are young and growing research group, working on the intersection of machine learning and computational systems residing on the edge, including embedded, mobile and distributed systems. We work on the development of theories and methods focused to running machine learning on resource-constrained devices, as well as designing system software to increase training and inference efficiency of machine learning computation on the edge.