Postdoc position in Computer Engineering at the Division of Computer Engineering, Department of Electrical Engineering, Linköping University.
Work assignments
The postdoc will carry out research directed by Prof Jose Nunez-Yanez, who is a Royal Society industrial fellow in Machine intelligence at the network Edge and has been appointed as Professor in Computing Architectures for Machine Learning at the department of Electrical Engineering (ISY), Linkoping University. The position is funded by WASP and it is expected that the successful candidate attends the annual workshops and presents the outcomes.
State-of-the-art deep learning models such as those used in object detection and machine health diagnosis are deploying new features such as attention mechanisms and data fusion which require a large amount of compute power and energy to run in real-time. One possible solution is to map them to cloud-based services, but this has negative implications on latency and data security. To address this challenge, we will research embedded deep learning heterogeneous computing platforms that combine hardwired, reconfigurable and programmable hardware. The heterogeneous approach can combine compute resources such as Google EdgeTPUs, Intel VPUs and FPGA logic to deliver optimal points of scalability, throughput, flexibility and low-latency. This will enable the deployment of complex models such as attention-based neural networks with recurrent layers or transformers to SoC health monitoring and fault diagnosis near data sources such as performance counters and hardware sensors. A complex multi-core RISC-V system could constitute the SoC under analysis for demonstration purposes.