Embedded Artificial Intelligence for Dark Matter Detection

Project Description

IMPETUS develops intelligent systems for systems requiring high-speed data analysis (>100 GB/s). These systems process, identify and compress data in real time using artificial intelligence algorithms embedded on or near detectors. Current projects include medical imaging, X-ray free electron laser (XFEL) experiments, dark matter detectors and intrusion detection in peripheral systems (cybersecurity). 
Normal matter only accounts for about 5% of the universe – the composition of the rest of the dark matter and dark energy in the universe remains a profound mystery. Identifying the dark matter, by constructing very sensitive detectors to search for rare interactions of dark matter particles with normal matter, is a top international scientific priority, with Canada playing a key role in the efforts to design and build ARGO, a 400-tonnes liquid Argon detector.
 
Objective 
 
The project aims to implement the algorithms built by the team for ARGO onto dedicated hardware. This involves further refining and compressing the models for hardware compatibility, designing the logical circuit and implementing it on hardware, designing the data flow systems and testing the system under realistic conditions to measure all performance metrics, including accuracy, power usage, latency and throughput. 
 
Methodology 
 
The project is being carried out with the Global Argon Dark Matter Collaboration. The Sherbrooke team is currently developing neural networks for particle identification and position, but these require further optimisation to make them functional on a real-time system. This optimisation involves reducing the network by means of quantification methods, pruning and, potentially, knowledge distillation. The student will use in house tools to convert the netwrok for FPGA implementation and measure the performance on our real-time DAQ testbench in the 3IT.micro platform.
 
Supervision and working environment 
 
The project will be carried out under the supervision of Pre Audrey Corbeil Therrien at the Institut Interdisciplinaire d'Innovation Technologique as part of the IMPETUS team. The successful candidate will interact regularly with all collaborators, with the possibility of a short stay with Canadian collaborators. The student will benefit from an exceptional research environment combining students, professionals, professors and industry working hand in hand to develop the technologies of the future. This project is supported by the McDonald Institute and the student will have access to their resources and activities. 
 
Specific requirements 
 
The candidate must have a good academic record, beginner knowledge of nuclear physics, intermediate skills in Python programming, a sense of creativity, a strong ability to adapt and a taste for research and development in instrumentation and artificial intelligence. 
 
Experience in neural network design, hardware design (FPGA) and/or astroparticle physics is an asset.

Funding offered

Yes

$ 25 000

The last update was on 24 October 2025. The University reserves the right to modify its projects without notice.