Imec-IPI is seeking a highly motivated and talented PhD Student to join our research team and to work in the field of Optimization and Adaptation of Deep Learning Systems to Edge and Cloud for Industrial Applications. This research opportunity is based on the urgent need for enhancing the efficiency and scalability of edge AI solutions in the industrial sector. You will be offered an initial contract of 12 months, to be extended up to a period of 4 years in total with the aim of obtaining a PhD. The research will be partly fundamental (furthering the state of the art) and partly applied. You will collaborate with international industrial partners while embedded in a university research team that is internationally recognized for its extensive expertise regarding sensor fusion for autonomous driving, traffic monitoring and industrial safety.
The Image Processing and Interpretation (IPI) group at Ghent University-imec consists of +/- 40 experienced researchers, post-docs and professors. IPI conducts research in a wide array of both fundamental and applied image processing topics. Application domains of this research include intelligent & autonomous vehicles, surveillance and sensor networks, remote sensing, medical image analysis, video analysis and scene reconstruction.
Ghent University consistently ranks among the best 100 universities in the world, including, 69th by the Academic Ranking of World Universities (or Shanghai ranking) and 88th by U.S. News & World Report. The IPI Lab is location on the university’s UFO campus in the center of Ghent, Belgium, a city recently rated as one of the best places to visit in Europe for culture.
You hold a Master of Science degree in computer science, electrical engineering, or equivalent. You possess a strong knowledge of mathematics, probability theory, image processing, machine learning, and computer vision and are well versed in Python and/or C++ (e.g., deep learning frameworks such as PyTorch or TensorFlow). You have experience with original algorithm design that goes beyond the mere application of methods from literature.
You combine a strong interest in engineering and scientific research with a desire to see your work applied (in industrial, academic or (N)GO collaboration). You are able to learn quickly and independently. You aspire to become an expert in your field, while simultaneously collaborating with other researchers and senior staff to efficiently generate state-of-the-art results.
You are fluent in written and oral English and able to communicate your original ideas and results clearly and concisely. You have excellent teamwork skills, and you are motivated to drive research innovation in the field of edge AI for industrial applications.
Opportunity to work on cutting-edge research in research projects in collaboration with industry and academic partners.
Access to state-of-the-art computing resources and facilities.
Guidance and mentorship from experienced researchers in the field.
The adoption of edge AI solutions in industrial settings has the potential to revolutionize manufacturing processes, quality control, predictive maintenance, and more. However, there are significant challenges to overcome in deploying and running deep learning models efficiently on edge devices. The industrial sector faces several pressing challenges in adopting edge AI solutions:
Limited Computational Resources: Edge devices in industrial environments typically have restricted computational resources, making it challenging to run large, resource-intensive deep learning models effectively.
Real-time Requirements: Many industrial applications require real-time or near-real-time inference, necessitating highly efficient model execution to meet strict latency constraints.
Scalability: Scaling AI solutions across a network of edge devices while maintaining consistent performance and resource utilization is a complex problem that requires innovative solutions.
This research project aims to address (some of) these challenges by focusing on three key aspects:
Performance prediction of existing neural networks: Developing novel methods to accurately predict the performance of deep learning models on edge devices. This involves analyzing and modeling various factors such as hardware capabilities, network architecture, expected throughput, latency and power consumption to optimize the model selection and adaptation.
Defining splits of neural network models: Investigating innovative techniques for dividing and distributing neural network models between cloud and single/multiple edge devices to balance computational loads and resource utilization. This includes dynamic model splitting strategies to adapt to changing conditions in real-time (e.g., malfunctioning edge devices) and requires minimizing the communication overhead when transferring data between devices.
Investigation of Message Passing Schemes for online training: distributed training using stochastic gradient descent (SGD) methods face big issues regarding communication overhead, data imbalances, dealing with hardware heterogeneity and complexity of the implementation. Recently, message passing algorithms based on belief propagation have shown to be a viable alternative to SGD, whilst offering more advantages when deploying these algorithms in a distributed context. The goal is then to design efficient message passing schemes taking the characteristics of the edge and cloud devices into account.
What we offer you
At IPI, we offer you the opportunity to conduct research in a highly international and friendly working environment. We provide ample opportunity for researchers to take initiative in their work and to develop their professional networks.
The salary is competitive and will be determined by the university salary scales. Staff members can count on a number of benefits, such as a broad range of training and educational opportunities, 36 days of vacation leave (on an annual basis for a full-time position), bicycle allowance, and more.
How to apply
Please submit your application by email to Prof. Bart Goossens with " [EdgeCloud] " in the subject line.
In your email, please include the following in 1 merged pdf file:
A cover letter with a brief motivation of your application: what do you consider the best aspects of your CV that demonstrate your academic excellence during your university education? What are your reasons for pursuing a PhD? Why would you like to work at UGent?
A detailed CV, describing your earlier experience and studies;
Contact details for three reference persons;
A list of publications (if available);
A transcript of your educational record (list of courses per year, number of obtained credits, obtained marks) if available. This need not be official documentation at this stage;
If available: 1-3 English language documents describing your earlier research or technical work (e.g., scientific papers, master thesis, report on project work, etc.). These documents need not be on the topic of the advertised position.
Applications remain welcome until the position is filled.