We are entering an era where Artificial Intelligence (AI) is embedded in every edge device, creating an urgent need for techniques to implement time-constrained applications on energy and resource-constrained edge hardware. As such, we are seeking a highly motivated individual to join our team and take on the challenge of developing the next-generation compilers for edge AI platforms with a particular focus on neuromorphic computing. This is a unique opportunity to work on cutting-edge technology and make a significant impact in the field of AI and edge computing.

Job Description

Cyber-physical systems (CPS) are a key enabling technology for Industry 4.0 (smart production, smart energy, transportation, etc.). CPS are a tight integration of computation with physical processes whose behavior is defined by both cyber and physical parts of the system. The computation part has stringent latency constraints that must be met to close the control loop and to guarantee the safety and stability of the system. Platforms that are used to run these computations do not offer timing guarantees.  To ensure that the timing constraints of the computation are not violated, designers, therefore, resort to over-dimensioning and overprovisioning of the resources assigned to the computation.

Neuromorphic computing is an emerging technology that offers the potential to address these challenges in a more efficient and effective way. By leveraging the principles of neural processing found in biological systems, neuromorphic computing architectures can enable ultra-low power, high-performance computing that is well-suited to real-time control and decision-making tasks in CPS. With the advent of edge AI, cyber-physical systems will become even more distributed than today, and the computational load of the algorithms running in CPS will increase drastically. Moving applications to the edge reduces latency and saves communication energy, making neuromorphic computing a promising technology for the development of energy-efficient and high-performance CPS platforms.



With the advent of edge AI, cyber-physical systems will become even more distributed than today, and the computational load of the algorithms running in CPS will increase drastically. Moving applications to the edge reduces latency and saves communication energy, making neuromorphic computing a promising technology for the development of energy-efficient and high-performance CPS platforms. However, these edge AI platforms have extremely limited compute resources and are often battery-powered increasing the need for energy-efficiency. A radically innovative approach is needed to map latency-constrained applications onto these novel edge AI platforms hereby guaranteeing that timing constraints are met without unnecessary resource consumption. In this project, you will research a novel mapping and design-space exploration framework that uses dataflow analysis techniques to study the timing behavior of energy-efficient applications running on edge AI hardware. The focus will be on minimizing the energy consumption of applications through innovations at the programming and system architecture level. To maximize the impact of these savings, innovative computing architectures such as in-memory computation, neuromorphic processors, and electro-photonic accelerators will be considered as target platforms for our framework. Your work will result in breakthroughs in the programming and design of distributed edge AI applications running on ultra-low power platforms through a programming flow that provides latency guarantees to applications when running on heterogeneous edge AI platforms. This reduces the overprovisioning of resources needed to meet the timing constraints of these applications. A prospective outcome of this project is that applications can run more efficiently on edge AI platforms allowing them to save energy. They can be applied to the development of Cyber-Physical Systems as used for example in the high-tech industry or in smart energy grids. Application in other areas with distributed computing (e.g., autonomous vehicles, patient monitoring) is also within reach.