PhD position on Privacy-Preserving Face Swapping on Embedded Hardware Powered by GenAI
The e-Media Research Lab at KU Leuven is offering a PhD position on privacy-preserving face swapping on embedded hardware powered by GenAI. The project sits within the STADIUS research section in the Department of Electrical Engineering (ESAT) and brings together signal processing, machine learning, data analysis, and human-computer interaction.
The research focuses on developing a real-time, privacy-safe camera system that anonymizes people directly on-device. Instead of simple blurring, the system will replace faces with natural-looking synthetic versions that preserve appearance cues such as gender, age, expression, and gaze while concealing identity. The raw video stream should never leave the device, and authorized reversal must remain possible when legally required. Potential application areas include privacy-preserving visual sensing in healthcare and anonymized person detection in surveillance and smart-city contexts.
The appointed PhD researcher will work on bringing face-swapping and de-identification algorithms to off-the-shelf embedded platforms. Key tasks include deploying the pipeline on resource-constrained hardware, optimizing AI models for latency, memory, and power usage, and building demonstrators for real-time anonymization. The project emphasizes hardware-aware neural architectures, hardware-in-the-loop optimization, and efficient inference strategies for edge-based processing.
The position is an excellent fit for a highly motivated candidate interested in optimizing AI on resource-constrained devices and working at the intersection of theory and implementation. The lab highlights its work in healthcare, Industry 5.0, biomedical sensing, and education, and is exploring tinyML technologies to improve intelligence and energy efficiency in constrained systems.
Applicants should have a Master’s degree in Electrical Engineering, be in the top 10% of their class in both MSc and BSc studies, and have exceptional grades. Strong deep learning knowledge, embedded hardware experience, solid programming skills in Python and C/C++, and familiarity with PyTorch or TensorFlow are expected. Interest in hardware-aware AI techniques such as model compression, quantization, pruning, and efficient neural design is important, as is fluent English and the ability to thrive in an international team.
The offer is a PhD scholarship for up to four years, subject to positive intermediate evaluations. Non-EER applicants may also receive up to one year of pre-doc support. KU Leuven provides an internationally recognized research environment, with opportunities to participate in conferences, workshops, and collaborations with top EU research teams. The position is planned to start from 1 October 2026, subject to discussion. The application deadline is 10 August 2026.
