Innovative Projects Realized

Learning Discussion Thread Representations to Empower Content-based Recommendation

VerticalScope is a company that owns online forums in many domains, such as automotive, health, technology, and powersports. VerticalScope uses a content based recommender system to mitigate the cold start problem, where a large portion of traffic on the forums are made by unregistered users. The goal of this project is to learn representations of discussion threads. Thread representations that capture semantic and contextual information can improve the recommender system to suggest more relevant threads to users, and boosts search engine optimization and user retention rate. Understanding user sentiments also allows for the discovery of trending topics, and personalized homepage and advertisements.
Learning thread representations has various challenges. Within the automotive forums for example, there may be multiple threads talking about buying and selling cars. However, though these threads may have similar context, the object of discussion (e.g. the specific car model) can be different, and the learned representations should capture these differences. Another related problem is that there are many out of vocabulary words that may be very important to the relevancy between two threads (e.g. the name of a specific product).

Goal-Conditioned Reinforcement Learning

The goal of the project is to improve upon the methodology behind goal conditioned learning. In this framework, similar to the setup in traditional reinforcement learning, an agent interacts with an environment. However, instead of training the agent to maximize return, the agent is trained to reach a given goal at the end of the trajectory. That is, given a rollout-specific goal, the agent attempts to reach it. This goal conditioned paradigm is particularly promising for applications where the objective changes in every episode, for example, controlling a robot or a drone for different tasks; or self-driving vehicles, where the destination might change between episodes. In this project, we will explore potential improvements within the goal conditioned framework, both in the discrete and continuous action space settings.

Développement d’un béton électriquement conducteur par utilisation des résidus de bauxite brutes ou transformés

L’objectif principal de ce projet consiste à développer une formulation de béton électriquement conducteur basé sur l’utilisation des résidus de bauxite calcinés.
Les retombées de ce projet de recherche pour le Canada sont doubles puisqu’elles permettront, dans un premier temps, une valorisation de produits recyclés issus de la production d’aluminium canadienne, réduisant ainsi l’impact environnemental de ces derniers. Dans un second temps, cela représente un potentiel économique non négligeable avec le développement d’un nouveau type de béton chauffant pouvant être utilisé dans plusieurs domaines en lien direct avec les conditions climatiques exigeantes du Canada tel que la protection hivernale des tabliers des ponts ou encore des pistes aéroportuaires qui requièrent une quantité importante de produits déverglaçants nocifs pour l’environnement et les structures.

Combined relational and BERT-ranking for multilingual ad hoc document retrieval

With increasing amounts of information available online on the web, it’s crucial for search engines to filter out the content they think is useful and rank that content in decreasing order of relevance to the user’s query so that the user can just focus on the top results. Traditional techniques in search ranking focused on presence of the user’s search terms in the documents being returned by the search engine. Now, modern advances in machine learning allow us to understand complex relationships between what the user really is looking for and what the documents really are about and this research will use this understanding to make better search engine rankings. This research also leverages these advances in technology to also understand the similarities between documents for return meaningful results to user search queries in multiple languages.

Net-shape Manufacturing of Fins for High Efficiency Heat Exchangers

Environmental pressure to reduce greenhouse gas emissions and fuel consumption has prompted
significant worldwide activity to find effective renewable/regenerative energy solutions. Solutions
where power is produced and waste heat from the exhaust is recuperated have great potential,
especially for distributed (decentralized) power generation (DPG). The latter approach reduces energy
losses that are caused by the absence of long-distance transmission lines and minimizes the risk of
widespread electrical failure. The development of high efficiency microturbines (MT) is critical to the
success of DPG and relies on efficient heat exchangers. Brayton Energy Canada has recently
developed a new proprietary design of high-efficiency heat exchangers that could allow MT to reach
higher thermal efficiency and thus quickly become a solution for DPG if parts (the fins) could be
manufacture in a cheaper way. The proposed project aims at developing a new manufacturing process
to produce the fins in a cheaper way.

Targeting SARS-CoV-2 (COVID-19) methyltransferases (nsp10-nsp14 and nsp10-nsp16 complexes) toward developing small molecule antiviral therapeutics

COVID-19 pandemic has brought the world to standstill with more than 3 million people infected and more than 200 000 mortality so far. It has literally brought the health care systems in many countries to the breaking point, if not beyond. The economic consequences have been devastating with millions of people out of work. We are taking a novel approach by focusing on two SARS-CoV2 (COVID-19) methyltransferases that are essential for viral replication. Both enzymes (nsp14 and nsp16) are druggable. Therefore, identifying potent inhibitor of these two proteins could be used in providing new therapeutics for COVID-19. In addition, because these proteins are highly similar in other coronaviruses such as SARS (SARS-Cov) and MERS (Middle East Respiratory Syndrome), the same drugs likely could be effective in treatment of other coronavirus infections.

Recommending Benefits Utilization to Promote a Healthy Lifestyle

Users on the League platform have access to a number of health and wellness benefits including massage, physiotherapy, personal trainers and a variety of other programs; however, not all of them fully utilize them to maximize their wellbeing. Utilizing the health and program utilization data we want to develop robust personalized predictions that will suggest to individuals, programs that they are eligible for and would benefit their health. We are hoping to further develop League’s platform into a health hub where every user will be promoted healthy behavior and wellness programs optimized for their health profile. Our recommendations will strive to make our users happier and healthier.

On the design of a new electro-refining process for the recovery of magnesium from Used Beverage Can (UBC) aluminum alloys

There is actually a societal challenge here in north America regarding the end-of-life management of used beverage cans (UBCs). China is no longer accepting several of our recyclable waste streams like UBCs. UBCs are made of aluminum that contains some level of magnesium and manganese to modulate the properties of the body and the lid respectively. UBCs can potentially be seen as a great secondary feed for the production of pure magnesium used to manufacture for example critical components in the automotive industry. It is evaluated that about 40000 tons per year of magnesium could be obtained from this waste feed stream.
The ultimate objective of this work is to design an improved electro-refining process technology to recover magnesium from aluminum melt. More specifically, we want to decrease the energetic requirements by lowering the electrode inter-distances and adjust the chemistry of the electrolyte to meet modern environmental standards.

Developing Intelligent BI Applications using Machine Learning

SAP is a multinational software corporation that makes enterprise software to manage business operations and customer relations. SAP Labs Montreal focuses on supporting the consumer products industry. It provides these industries with the software infrastructure that allows its customers to run end-to-end processes including the capturing and analysis of sales transactions. Sales transactions include for example purchases made by each and every customer of the retail industry. Sales transactions are generated by point-of-sale (POS) systems used by cashiers and sent to a central server for validation and consistency checks. Recently, SAP has been actively applying machine learning to improve the efficiency of its software products. One challenge faced by development teams in this process is the identification of relevant features from raw data and the testing of the models. Current existing techniques are semi-automatic and often labor intensive. They also depend heavily on the expertise of domain experts, which often constitutes a significant barrier to the pursuit of machine learning projects in the Industry.

Towards a lab-on-a-chip-based rapid-screening system for pathogens

In light of the recent outbreak of Covid-19, it is urgent to find a solution for quick and efficient pathogen detection and elimination. Rapid point-of-need diagnostic tests and monitoring devices are urgently required in order to provide testing and care to those infected. Currently, testing is performed at centralized facilities using specialized equipment for molecular-based pathogen detection. Real time-quantitative PCR is the current method for detection, but it has a slow response time due to clinical lab capacity and sample shipping time.

In an effort to rapidly examine cells and microorganisms, we are developing an embedded sensor for lab-on-a-chip platforms with connected micro-tubes and a container for markers and the specimen. The sensor has built-in energy-harvesting and bidirectional communication units to create a contactless platform and analytical support for lab-on-a-chip technology. The proposed sensor attached to a microfluidic capillary carrier or lateral flow-based assay, facilitates rapid analysis and detection of harmful pathogens, drugs or biomolecules. It could also be used as a low-cost point-of-care patient monitoring device.

ChainCert platform: Verification authentication platform using Blockchain network

Blockchain is a well-known immutable, transparent and anti-counterfeit platform. As such it can be appropriate for submitting certificates for validation and safe-keeping. However, because of its requirements for complex and heavy computations for enhanced security and verification of each submission in the network, in this proposal, we have decided to conduct an experimental study to firstly evaluate several possible Blockchain candidate technologies to determine the ones that are the most cost-efficient and have a better performance for our case study. Next, we will design a model based on Blockchain network to validate the authenticity of each certificate that has been issued. The proposed method has a great impact to make sure that the paper-based certificates are authenticated and original.
This project has been suggested by the organization.

A 20 minute molecular diagnostic of COVID-19 within a “swab”

The research project will find optimal conditions for instrument-free, fast, and sensitive/specific detection of pathogen (COVID-19). The project is composed in two units(IU), each module within IU is offering independent technical solution for the current bottleneck in diagnostics industry. The first module is temperature-based denaturation of biological sample, coupled with nucleic acid-based detection (IU1). The second module is composed of specific amplification of biomarker for intended target/s, (instrument free) and the last module is colorimetric detection of amplified biomarker. These steps are essential for building universal diagnostic device which is projected to haveas high quality performance, as current state-if-the-art techniques, but is also fast (20 minutes) and does not require any additional technological setting. The partner organisation will use the technical solution provided by this project and generate it’s own, in-house, nucleic acid isolation/purification/detection techniques.