Projets novateurs réalisés

AI

The project intends to create AI agents for the financial services and insurance, transportation & logistics and public sector verticals. This will allow us to expand the reach and adoption of AI and GenAI at the Enterprise by adding ML capabilities as part of the intelligent automation solutions we build.
The candidate will be responsible for performing requirements discovery, exploratory data analysis, model development and model deployment.
The approaches to be utilized will range from statistical modelling, time series, Natural Language Processing, Computer Vision to LLMs, depending on the specific requirements of the solution.

Required Skills :Masters or PhD students in Computing Science, Mathematics, Physics, Engineering or related fields, focusing on Data Science would be the preferred profile. Experience in development of machine learning solutions, including NLP, Computer Vision, Generative AI, Time Series and knowledge of Python. Experience with cloud environments, like Microsoft Azure, Amazon AWS or Google GCP and the associated ML/AI suites. Knowledge of Vector Databases, Retrieval Augmented Generation and Large Language Models is an asset.

Towards Accelerating the Social Tipping Points in Sustainable Behaviors: A Behavioral Segmentation of the Quebecer Population

The societal ecological transition required to address climate change must occur rapidly, to save as much as we can of the planet’s ecosystem. To do that, long-term moralization processes need to be complemented with short-term behavioral change interventions.
A key component of successful behavioral change, whether in the short or long run, is understanding the target audience, recognizing that individuals differ in their motivation, intentions, and most importantly in their readiness to change. Therefore, a segmentation exercise is primordial.
This project will conduct a segmentation through a representative survey of approximately 2,000 Quebec residents, guided by the behavioral change framework, and the long-term moralization processes mentioned above. The segmentation will be theory-driven, incorporating behavioral change theories as well as moralization theories. The analysis builds on both hierarchical and non-hierarchical clustering methods for segmentation. This approach ensures theoretical relevance, generalizability, and empirical validation.
The objectives are to develop a comprehensive guide for designing targeted interventions within Quebec, establish a theory-driven segmentation model, and create a framework bridging short-term situational changes with long-term moralization. The results will inform policy and intervention design, facilitating a faster and more effective ecological transition.

L2M – A high-sensitivity, field-portable laser-based probe for soil monitoring

We are developing a portable soil analysis system based on laser-induced breakdown spectroscopy (LIBS). During LIBS, an intense laser pulse generates plasma on the soil surface, and the emission from this plasma serves as a fingerprint of the soil’s elemental composition. The signal variability caused by the heterogeneous soils and challenges of quantifying complex composition are overcome by machine learning (ML) data modeling. This novel ML-LIBS device will significantly reduce the cost and processing time of soil analysis, supporting real-time monitoring of soil nutrients, composition, and fertilizer optimization for sustainable cultivation.

Grizzly Bear Movement Ecology, Habitat Selection, and Population Connectivity between threatened populations in southern British Columbia

This research project is focused on grizzly bear populations in southern British Columbia at the extent of their
species range. While population recovery has occurred in parts of the ragnge, some popualtions continue to deline
and large areas remain extirpated. Where the degree of population fragmentation is known this project will focus
on analyzing how bears choose their habitat, their diet, and behavioural patterns are affected by hydroelectric
development in key areas. Where the degree of population fragmentation is less known the project will use genetic
information to understand how populations are connected and what landscapes and human infrastructure help or
hinder connectivity. The ultimate aim is to give Indigenous communities the tools to lead conservation efforts and
guide grizzly bear recovery on their lands. This work is deeply important, as it respects and supports Indigenous
stewardship of the environment.

Estimation de la variance des estimateurs semiparamétriques de l’efficacité vaccinale avec le devis test-négatif : étude de simulation.

The test-negative design (TND) is an observational study design that is currently and routinely being used globally to evaluate the effectiveness of vaccines against COVID-19 illness caused by emerging SARS-CoV-2 variants. Recently, more reliable statistical estimators were proposed to get better estimates of vaccine effectiveness. However, no one has yet investigated what statistical methods will best allow us to understand the accuracy of these estimates (which is a fundamental component of statistical analysis). This project proposes to compare different ways to estimate the estimator accuracy using synthetic data generated under a hypothetical TND.

Using the Arabidopis toolbox to evaluate the plant growth promoting activity of purified molecules from brown algal extracts

Seaweeds and seaweed products have been promoted in agriculture as source of nutrients and activators, to improve plant growth, plant productivity and food production. A wide range of beneficial effects have been observed, including seed germination, enhanced growth and crop yield, elevated resistance to biotic and abiotic stress. However, the bioactive compounds have not been identified using classical methods of bioassay-guided fractionation and the mechanisms of action remain poorly understood. The aim of this exchange project is, at first, to test purified fractions of commercial seaweed extracts that are candidates to promote plant growth using rapid bioassays that were develop by the host laboratory to evaluate the plant growth promoting activity on the model plant Arabidopsis thaliana. These tests will complement a study of the physiological effects of these seaweeds extracts in the model plant Arabidopsis thaliana through transcriptomic and metabolomic approaches that is conducted within the PhD thesis of the applicant.

Structural assessment of Eastern Canada’s transitional unreinforced masonry buildings: case study analysis

This research project will contribute to the preservation of the existing building stock and mitigate potential seismic risks by developing simplified numerical models to contribute to the seismic assessment of local unreinforced masonry (URM) buildings. Old URM structures are a prominent building type across Eastern Canada, contributing to the rich architectural heritage and cultural fabric of the region. Despite their vulnerability to earthquakes, and as Eastern Canada is a moderate seismicity region, the seismic assessment and strengthening of old URM buildings remains an important yet understudied focus. The focus of this research project is to develop accurate numerical models for the seismic analysis in the case study of an old URM industrial building, typical of Québec, using two novel, simplified modelling strategies implemented in two commonly used software for structural analysis. The first technique is by using a macro-modelling strategy in 3DEC, a distinct element modelling software widely used for masonry. The second modelling strategy uses a macro-element developed for analysis of URM in OpenSees. Results expand the understanding of the behaviour of typical Eastern Canadian buildings, important to inform future assessment and retrofit procedures.

Accessible Communications of the Needs of Persons with Disabilities

Building codes are often written in complex legal language, making them difficult to understand. While diagrams can help, they often lack standardization and fail to convey the intent of the requirements. To help architects, engineers, contractors, and inspectors to understand the intent and specifics of requirements, we aim to develop a set of best practices for creating successful instructional diagrams. After completing a baseline study of existing diagrams of accessibility requirements, we will propose a set of accessible diagram principles for both the creation and automated compliance testing of diagrams. We will then apply those principles and create and test an open source library of validated visual learning components for building codes. This work will benefit the construction industry overall and Trax in particular to augment the licensed codes and standards in the Trax Codes product.

Magnetoelastic coupling in novel quantum spin liquids

A quantum spin liquid (QSL) is a type of magnetic material in which the spins of the electrons are highly quantum entangled. This phase of matter has particularly exotic quasi-particles – ways of increasing the energy of the system that resemble quantum particles, but also have novel properties as compared to those we are accustomed to observing in our universe. In some cases these quasi-particles have properties that make them potentially useful for encoding quantum information. Discovering and understanding new QSLs is one of the major challenges of the field of condensed matter physics and has the potential to reshape our approach to technologies such as quantum computing. In this project, the intern will travel from Université Paris-Saclay to the Université de Sherbrooke to carry out sound velocity measurements on potential QSL materials. The goal is to understand how their quasi-particles couple to the crystal lattice, to assess in what ranges of magnetic field and temperature spin entanglement is present and to complement his work at Paris-Saclay using magnetic resonance on related materials.

Assessing the exposure of Arctic terrestrial ecosystems to global changes via animal migration

Every year, millions of migratory birds travel to the Arctic to breed during the summer, then migrate to different parts of the world in autumn to spend the winter. These major seasonal flows of animals transport large quantities of energy, nutrients, contaminants and pathogens, across the globe. It is therefore essential to determine the migratory destinations and the environmental conditions that Arctic migratory species face outside of the Arctic to effectively assess the exposure of Arctic ecosystems to global changes. Our objective is to assess the exposure of an Arctic ecosystem to environmental changes occurring in remote, but connected environments through animal migration. Using a network approach, we identify the migratory connections between an Arctic breeding ground and hundreds of ecological regions of the globe used as non-breeding grounds. Subsequently, we assess the exposure of each species to environmental changes on their non-breeding grounds using global maps. The project is conducted in collaboration with the University of Québec at Rimouski (UQAR) and the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL). This project aligns with UQAR’s major research focus on northern studies and supports the mission of WSL by improving our understanding of global ecological dynamics.

Virtual Hazard Perception: Distinguishing Novice Drivers by Experience in a VR Training Environment

The proposed research explores how a virtual reality (VR) driving simulation can assess drivers’ ability to perceive hazards and how accurately it reflects their real-world driving experience. Participants from UBC will engage in a lab-based VR simulation where they will navigate various driving scenarios with the goal of identifying potential road hazards. Their performance will be scored, and the data will be analyzed, to determine if the VR simulation can accurately predict the amount of driving experience participants have. This project will help Project Whitecard, the partner organization, assess the potential of VR-based training tools for driving safety and education providing valuable insights for future driver training programs across Canada.

L2M – BackSCNR

BackSCNR offers an accessible solution for screening and monitoring AIS designed to provide sustainable care for patients. By simplifying and shortening medical appointments, BackSCNR reduces the burden on patients and the healthcare system while also offering patients an easy way to track changes in their spine curvature over time. The technology behind BackSCNR is supported by over a decade of research and has secured numerous competitive research grants. Researchers and clinicians from countries including Canada, Brazil, Spain, Chile, and Egypt have contributed invaluable insights to this project and tested the technology on over 1,000 participants.