The cloud-to-edge continuum, comprising a variety of computational resources from cloud data centers to resource-constrained devices at the network edge, offers many possibilities to optimize the deployment of machine learning (ML) applications. Different resources in the continuum may be associated with different advantages, e.g., cloud data centers offer practically unlimited compute capacity, whereas edge devices potentially offer low-latency communication with data sources, some nodes may offer special hardware optimized for ML operations (GPU, TPU), etc. Based on the needs of a specific ML application, network parameters, and the characteristics of available devices, different deployment strategies may be preferable, such as a cloud-only deployment, a deployment to a single edge device, a deployment over a set of edge devices, etc. The deployment may also be changed during the operation of a machine learning model, e.g., performing training in the cloud and inference at the edge.
Existing approaches to the deployment of ML in the cloud-to-edge continuum mainly focus on two aspects: maximizing accuracy (i.e., some measures for the ratio of correct inferences made by the ML model) and minimizing latency (i.e., the time needed to train the model or to perform inference). Optimizing other aspects of ML in the cloud-to-edge continuum is far less common. Two of the most prominent additional aspects that could be considered are:
- Energy consumption is crucial for two reasons. First, ML applications can consume large amounts of energy, leading to high costs and high environmental impact (carbon footprint). Second, energy consumption is especially critical for battery-powered edge devices.
- Information security is crucial to ensure that the processing and transfer of data does not violate confidentiality or privacy requirements. For example, if the training data or the inputs for inference involve personal data, this data must be protected in accordance with the GDPR.
Considerations of energy consumption and information security may impact deployment decisions, since different deployment options may have different influence on energy consumption and on information security. In addition, energy consumption and information security considerations may conflict with each other or with the more traditional accuracy and latency considerations. For example, processing data in an edge device may entail low latency, but may not be feasible because of the limited energy budget of the device, while moving the data to another device without energy limitations solves the energy problem but may lead to a security threat, which could be mitigated by anonymizing the data, but this may lower accuracy etc.
What are you going to do
You will develop new approaches for determining the optimal (or near-optimal) deployment of ML applications in the cloud-to-edge continuum, with a focus on energy and security considerations, while also accounting for accuracy and latency. Potential sub-goals:
- Determining the set of possible deployment options for an ML application, considering different types of devices (e.g., spreading the layers of a deep neural network across a set of edge devices), and different security controls, such as encryption, anonymization, obfuscation, trusted execution environments;
- Determining appropriate metrics for quantifying the energy consumption and information security implications of different deployment options;
- Determining appropriate models for capturing the requirements associated with a ML application and the capabilities and constraints associated with the cloud-to-edge continuum;
- Developing an algorithm for determining a set of Pareto-optimal deployment options for a given ML application;
- Consideration of multi-domain/multi-owner aspects of deployment scenarios.
The ultimate demonstrator for this project could be a model that can serve as a digital twin for such infrastructures, enabling informed decision making on a complex combination of risks and benefits with real-time adaptive response. Here, data access and sovereignty requirements can take extreme forms, which can lead to inconvenient and expensive solutions, like complete network separation. Embedded in the Complex Cyber Infrastructure group (CCI), you will become part of a large group of fellow PhD students and researchers working on connected topics.
What do you have to offer
Your experience and profile:
- An MSc degree in computer science or a related discipline;
- Enthusiasm for the research process: studying research papers, solving complex problems, applying creative thinking, evaluation, reflection and disseminating findings via writing and oral presentations;
- The ability to design modular and scalable software systems with reference implementations;
- The ability to critically analyze abstract models as well as concrete implementations;
- The ability to work individually as well as effectively in a team;
- Fluency in written and spoken English.
CONDITIONS OF EMPLOYMENT
Fixed-term contract: 18 months.
A temporary contract for 38 hours per week for the duration of four years (the initial contract will be for a period of 18 months and after satisfactory evaluation it will be extended for a total duration of four years). This should lead to a dissertation (PhD thesis). We will draft an educational plan that includes attendance of courses and (international) meetings. We also expect you to assist in teaching undergraduates and master students.
The gross monthly salary, based on 38 hours per week, ranges between €2,443 to €3,122 (scale P). This does not include 8% holiday allowance and 8,3% year-end allowance. A favourable tax agreement, the ‘30% ruling’, may apply to non-Dutch applicants. The Collective Labour Agreement of Universities of the Netherlands is applicable.
Besides the salary and a vibrant and challenging environment at Science Park we offer you multiple fringe benefits:
- 232 holiday hours per year (based on fulltime) and extra holidays between Christmas and 1 January;
- Multiple courses to follow from our Teaching and Learning Centre;
- A complete educational program for PhD students;
- Multiple courses on topics such as leadership for academic staff;
- Multiple courses on topics such as time management, handling stress and an online learning platform with 100+ different courses;
- 7 weeks birth leave (partner leave) with 100% salary;
- Partly paid parental leave;
- The possibility to set up a workplace at home;
- A pension at ABP for which UvA pays two third part of the contribution;
- The possibility to follow courses to learn Dutch;
- Help with housing for a studio or small apartment when you’re moving from abroad.
Are you curious about our extensive package of secondary employment benefits like our excellent opportunities for study and development? Take a look here.
University of Amsterdam
The University of Amsterdam is the largest university in the Netherlands, with the broadest spectrum of degree programmes. It is an intellectual hub with 39,000 students, 6,000 employees and 3,000 doctoral students who are all committed to a culture of inquiring minds and scientific excellence.