Innovative Projects Realized

Development of fairness-aware training and fine-tuning methods for tabular foundation models

In today’s world, organizations such as banks, schools, and hospitals rely on artificial intelligence (AI) to help make important decisions, from assessing credit risk to detecting fraud. A new type of AI, called Tabular Foundation Models (TFMs), shows great promise for these kinds of tasks. Unlike traditional systems that require heavy retraining, TFMs can quickly adapt to new situations using only a small amount of data, making them faster and more flexible. However, before these models can be trusted in high-stakes settings, it is essential to ensure they make fair and unbiased predictions. Biased AI systems can unintentionally discriminate against certain groups of people, leading to unfair or harmful outcomes.

This project will focus on building fairness directly into TFMs so they remain accurate while also treating all groups fairly. For Layer 6 AI at TD Bank, this research will provide practical methods and tools to build more responsible AI systems. More broadly, the results will contribute to advancing fair and trustworthy AI in Canada.

Characterizing the three-dimensional architecture of the tumor microenvironment in prostate cancer bone metastasis

Prostate cancer is the most common male cancer in Canada. While many patients live long lives with treatment, some develop incurable disease that spreads to the bones. When this happens, survival drops sharply and patients often face severe pain, broken bones, and serious complications such as spinal cord damage. Bone changes caused by prostate cancer can look very different: sometimes the bone is eaten away, and other times it becomes overly dense but weak and fragile. Current treatments help slow bone loss and manage pain, but they do not stop the abnormal bone growth or the processes that drive pain. Research shows that prostate cancer cells change their surroundings in bone. They encourage the growth of new blood vessels to feed the tumour and new nerves that both fuel cancer growth and make bones more sensitive to pain. So far, drugs aimed at stopping these changes have not worked well in patients. Our project will use advanced 3D imaging on donated patient bone samples to see exactly how cancer cells, bone cells, blood vessels, and nerves interact. This new understanding could point to better treatments that reduce pain and improve life for patients with advanced prostate cancer.

The Business Case for Sustainability

This project explores how the sustainability efforts of select Atlantic Canadian businesses are enhancing their competitiveness and profitability, as well as positively affecting the communities and systems they operate within. We will develop 10-12 mini case studies, highlighting how organizations’ actions have strengthened their business model and competitive advantage while simultaneously creating social and/or environmental benefits. Our purpose is to raise awareness of the positive effects of businesses committed to sustainability and to offer clear, tangible and feasible steps for others to take in this direction.

Évaluation rétrospective de l’impact du positionnement sur le sommeil des grands prématurés

Après la naissance d’un grand prématuré, le développement de son cerveau se poursuit à l’intérieur de l’incubateur. Favoriser ce développement de façon optimale est une priorité des soins, en particulier pour les infirmières néonatales. L’une des stratégies pour y parvenir est de promouvoir un sommeil de qualité, avec le moins d’éveils possible. Le sommeil est une période essentielle pour créer de nouvelles connexions neuronales et renforcer celles déjà présentes. Le positionnement des nouveau-nés dans les incubateurs semble avoir un effet important sur la qualité du sommeil. Aujourd’hui, trois positions standards sont utilisées en néonatalogie : dorsale, latérale et ventrale. Toutefois, la complexité des mesures du sommeil empêche de formuler des protocoles de soins précis quant à la fréquence d’utilisation de ces positionnements. Ce projet de stage à la Stanford University, vise à tester un nouvel outil : un algorithme de détection automatique du sommeil pour obtenir des données précises sur l’influence de chaque positionnement sur la qualité du sommeil des nouveau-nés prématurés. L’objectif de ce stage est d’obtenir des données rigoureuses, auprès d’une large cohorte de grands prématurés, pour mieux comprendre comment le positionnement dans l’incubateur influence le sommeil, afin d’orienter les soins et faire progresser la recherche néonatale.

Evaluating Mushroom-Derived Chitosan for Biocompatibility and Printability in 3D-Printed Medical Implants

Current implantable biomaterials often rely on synthetic or animal-derived polymers, which can be costly, poorly biocompatible, or environmentally unsustainable. There is growing interest in finding plant-based or fungal alternatives that meet the demanding requirements of medical-grade applications. Chinova Bioworks, a Canadian biotechnology company, specializes in producing chitosan from mushrooms for use in preservative applications. The company looks to expand into the medical sector by evaluating the potential for their chitosan as a base material for 3D-manufactured implants. MaterializeBio, a U.S.-based biomedical firm, focuses on developing and testing bio-compatible scaffolds for medical use, their challenge is to find new, safe, and effective biomaterials that can be integrated into their biomanufacturing platforms. This four-month project addresses both partners goals by assessing the physical, chemical, and biological properties of mushroom derived chitosan. The student researcher will conduct a literature review, lab testing (e.g. molecular weight, viscosity, printability), and basic biocompatibility assessments in collaboration with both organizations. This project enables Chinova to explore the medical-grade application of their mushroom-derived chitosan, potentially opening new markets in medical devices. Furthermore, the project provides Materialize Bio with data and insight on a potentially new, bio-compatible material that could be incorporated into their existing or future biomedical platforms.

Participation et bien-être des utilisateurs du bras JACO

Le bras robotisé JACO de Kinova Robotics® fait l’objet de recherche en continu afin de couvrir toute l’étendue des connaissances sur ses impacts sur les utilisateurs au Québec. Ce projet se donne comme objectif de réaliser une analyse secondaire d’entrevues précédemment menées dans la province de Québec auprès de 21 utilisateurs du bras robotisé JACO afin de décrire leur niveau de participation à toutes sortes d’activités et d’identifier les impacts de l’utilisation de cette technologie sur leur bien-être. Les résultats de cette analyse permettront d’élargir nos connaissances concernant les impacts de cette technologie en matière de participation et de bien-être des utilisateurs de cette technologie.

Investigation of state-of-the-art tools and methods for analysis of PTSD using neurovisualization

This project investigates existing and potentially new approaches to analyze functional brain structures in individuals with post-traumatic stress disorder (PTSD), using advanced neuroimaging such as fMRI, and machine learning. The goal is to identifying patterns built on dependent features within the images representing the connectivity matrices of brain regions. The project will investigate whether causal AI approaches such as Granger causality, as well as probabilistic graph models such as Bayesian networks can be used to analyze probabilistic dependencies between brain regions.

Making the Invisible Visible – Translating Groundwater Science and Engineering to Ensure Canada’s Water Security

One third of Canadians drink groundwater, yet the threats to groundwater remain ‘out of sight, out of mind’ for most Canadians and decision makers. Small, rural communities and First Nations disproportionately rely on groundwater as a drinking water resources but often lack the resources to advocate for effective policy that ensures groundwater security. Although groundwater is Canada’s largest freshwater resource, it is significantly underrepresented in national policy and management frameworks.

The MITACS intern will work between Dalhousie University and the International Association of Hydrogeologists – Canadian National Chapter (IAH-CNC) to conduct applied research on effective means of translating relevant groundwater science into educational and policy material. Key knowledge translation and science mobilization priority initiatives of the IAH-CNC will be advanced through the associated policy and management products produced. The intern will work closely with senior hydrogeology consultants, government professionals, and academics within IAH-CNC membership to develop and refine groundwater management or policy recommendations, conduct stakeholder meetings, develop groundwater communications material, prepare an opinion editorial or review paper, and attend/lead key water-related conferences events to strength networks.

The search for novel enzymes capable of biodegrading polystyrene, polyethylene, and polyvinyl chloride

Petroleum-based plastics are recalcitrant and pervasive environmental contaminants. Recycling only recovers 9% of post-consumer plastic, the rest is landfilled, incinerated, or discharged into the environment. While most plastic recycling methods convert plastic into lower value products, biocatalytic plastic depolymerization to component monomers then available for resynthesis of valuable products is an attractive alternative. Biocatalyst-based recycling is already used to recycle polyethylene terephthalate (PET). However, we do not have promising enzyme candidates for recycling for the most abundant plastics: polyethylene, polystyrene, and polyvinyl chloride which collectively account for ~45% of plastic waste produced annually. The homoatomic C-C backbone of these polymers renders them more resistant to degradation than the heteroatomic backbone of plastics like PET. Enzymes capable of oxidatively cleaving C-C bonds are of particular interest in development of biocatalyst-based recycling of these plastics. Laccases are a key enzyme family that cleave C-C bonds and that have the potential to impact C-C backbone plastics. The overarching goal of the Mitacs Globalink research project is to identify novel laccases by synthesizing and screening functional metagenomic libraries from diverse environments, and preparing these novel laccases identified for expression and purification to assess their suitability as biocatalysts for industrial recycling of C-C backbone plastics.

L2M – PORT-EM: Smart Port Energy & EV Infrastructure Management

The transition to low-carbon ocean economies requires innovative energy solutions for ports and coastal communities, where electrification of ships, ferries, and vehicles is accelerating. Building on prior research in probabilistic EV charging demand forecasting, energy consumption modeling, and emissions analysis, this project proposes a Smart Port Energy and EV Infrastructure Management Platform. The system integrates machine learning algorithms (LSTM, Random Forest, SVR etc.) to forecast highly variable port energy demand while incorporating tidal, offshore wind, and wave energy generation models. Coupled with embedded controllers for real-time load management, the platform enables demand balancing, peak reduction, and emissions minimization. Unlike existing generic energy management tools, this solution is purpose-built for the multi-source, high-variability environment of maritime ports, directly linking EV adoption with the ocean economy. The project will validate both the technical feasibility and market potential of ocean-focused smart grid solutions, advancing sustainable port operations in Atlantic Canada and beyond.

L2M – Medication Adherence Device

Older adults have difficulty keeping track of the frequency of their dosage as well as opening the medication packaging. Nonadherence to prescribed medication can lead to hospitalization which increases the burden of care both for the patients and the medication system. Our solution, Pillsette, is a smart medication adherence device that offers a reusable pill storage unit that can be operated manually or automated, and a companion app or SMS agent that alerts the patients (or their caregivers) at the time for the medication.

Jumeau Numérique: Profil d’échange thermique de la cuisson du sucre à la crème

NASUCO CANADA est une entreprise québécoise qui se spécialise dans la production, la transformation et la commercialisation de petits fruits, notamment des bleuets sauvages. L’entreprise souhaite optimiser son procédé de fabrication afin de soutenir sa croissance et améliorer la qualité de ses produits. Un des besoins essentiels de l’entreprise est d’optimiser le temps de cuisson du sucre à crème.
Le système de contrôle thermique automatisé (PID) actuellement en place repose exclusivement sur la mesure de la température. Cette approche, limitée à une plage restreinte de températures, ne tient pas compte de l’apport calorifique spécifique de la vapeur sous pression dans le processus de cuisson. Face à cette contrainte, plusieurs entreprises du secteur contournent le dispositif de régulation en utilisant l’interrupteur de pression de sécurité comme mécanisme de contrôle. Bien que cette pratique permette une amélioration ponctuelle de l’efficacité, elle engendre une grande variabilité du procédé et pose d’importants enjeux en matière de santé et sécurité.

Nos observations sur le terrain démontrent par ailleurs l’existence d’un profil de température particulier lors de la cuisson du sucre à la crème. En effet, les phases d‘homogénéisation des ingrédients, d‘ébullition de l‘eau et de cuisson du sucre vont chacune influencer, le taux de gras, de sucre et d‘eau dans le mélange, ce qui impact directement la chaleur spécifique du sucre à la crème. Lorsque le profil de chaleur spécifique change, la sensibilité de la pression à l‘intérieur de la double parois de la marmite passe d‘un état stable à un état instable en quelques secondes, rendant les paramètres de commande standard inadéquat dans une situation ou dans l‘autre, ce qui explique les limites d’un système de régulation basé uniquement sur la température pour assurer une gestion optimale et stable du procédé.