Projets novateurs réalisés

Characterization of Silicon Pixel Detectors for the High-Luminosity Large Hadron Collider

Development of a robust SPR system for the detection of growth factors in an insect-based expression platform

Growth factors are signaling molecules used in cell culture that provide cues to cells. They are one of the key inputs and main consumables in lab growth meat production. Because of this, they are the main driver of production costs. Future Fields is a Canadian biotechnology company that has developed a more cost-effective and scalable platform for producing key growth factors for lab grown meat production; however, analysis of growth factors produced in Future Fields’ system is costly and time consuming. Surface plasmon resonance (SPR) is a novel method of efficient and highly sensitive detection of proteins. The implementation of SPR into Future Fields’ production system would have profound impacts on their ability to produce growth factor products faster and more cost effectively.

Persuasive technology – Guidance to Virtual Relationship Manager (VRM) for effective sales effort basis voice data mining

Voice of the Customer (VoC) is how companies hear and listen to customer feedback about their brand, products, and services. Voice of the Customer solutions convert gathered feedback into valuable data and insights at scale. Data-driven VoC analytics programs are proven to increase customer lifecycle value and lower customer churn. Companies in various industries including insurance, financial services, and healthcare are leveraging this technology to generate insights into customer needs. ICICI Bank has large number for Virtual Relationship Managers who engage with customers assigned to them through electronic channels only for various product sales and servicing. This project’s primary objective is to develop a machine learning model which incorporates both customer record available with the bank along with voice data for finding potential buyers of the products of the bank. We also aim to create a voicebot to make personalized recommendations based on customer history and preferences.

Asset Vision: Extracting Metadata from Engineering Drawing Images

Organizations have traditionally struggled with asset management as a result of not having a complete picture of the location and state of their physical assets. This is largely because critical information such as how an asset was built and should operate is locked in highly technical diagrams and unstructured documents. Asset Vision is a new machine learning solution that automates the asset tag extraction process yielding critical asset information in a more time and cost-effective manner than hiring professional experts. The solution applies techniques from Optical Character Recognition, Natural Language Processing, and Machine Learning to identify, classify, and extract asset information from documents.

Multiplexed Source of Photons for Next Generation Quantum Secure Networks

With quantum computers which are just starting to hit the market, today’s encryption methods will soon become obsolete. In order to ensure an unconditionally secure data storage/encryption, communications will have to rely on a technology making use of the laws of quantum mechanics. The present MITACS application proposes to build the prototype of an optical source specifically designed for quantum secure communications as well as the development of quantum networks. This proposal represents a significant step towards the realization of compact, stable and energy efficient sources, thus addressing two major issues in quantum communications i.e. scalability and flexibility. Thanks to this MITACS application and stemming from the fast evolution of the quantum market which demonstrates the great efforts made by the scientific community to bring quantum applications to the market, we intend to push forward the development of practical photon sources.

Data Science and Machine Learning Algorithms for Event Sequence Data

Everyday millions of customers move through the sales cycles of companies, generating numerous data for potential use. The main objective of the research project is to advance the current state of the art techniques inside the company, with respect to the application of new algorithms on customer behavior data. From a research perspective, access to large sets of complex real world data can enhance great possibilities to apply and evaluate existing techniques at scale and to develop exciting new ones. This research project will involve developing, improving and implementing algorithms that provide advanced analysis of large sources of event data. In addition, the research projects will contribute directly and immediately to ODAIA’s range of products.

Generating Personalized Exercises in an Intelligent Tutoring System

Intelligent tutoring systems (ITS) are computer programs powered by artificial intelligence that can automate the tutoring process. They have the potential to provide low cost and highly scalable one-on-one tutoring to students around the globe in real-time. Their continued improvement is exciting for the future of education, and should be encouraged. Recent years have seen advances in the personalization of each student’s learning experience with an ITS (Kochmar et al., 2020). To further research into personalization, these systems should have the ability to provide question versions that are customized to the needs of each student. Pedagogical studies show that asking questions in the right way can hugely improve the learning gains of the student (Ashton-Jones, 1988) (Hrastinski, 2019). Generating question variants that can better match the needs of different students is the first step to help ITS ask each student the right questions, which is the goal of this project.

Video Spatial Recognition

Autonomous unmanned aerial vehicles (UAVs) are receiving significant attention in many communities, including academia, industry, and consumer electronics. SOTI is the world’s most trusted provider of mobile and IoT management solutions and its new aerospace division, SOTI Aerospace is focusing on hardware and software systems to support self-navigating situationally aware aerial drones. This project belongs to SOTI aerospace division and focuses on a vision system for the indoor environment. The primary objective of this project is finding out a real time 3D object recognition methodology to support drone navigation. Initial applications will be focused on the medical sector and search & rescue operations.

COVID-19 and resiliency: How Registered Practical Nurses working in long-term care adapt in times of personal, professional, and institutional crisis

Working in partnership, interdisciplinary professionals and the Registered Practical Nurses (RPNs) Association of Ontario (WeRPN) aim to understand what contributes to and detracts from personal, professional and institutional resiliency for RPNs during COVID-19. This is a rare opportunity to learn from front-line workers directly during historical social periods, such as COVID-19, and even less often does the public become acutely aware of the residents’ social inequities within the long-term care (LTC) sector and the indispensable nature of the largest regulated LTC workforce. This partnership will research ongoing first-hand experiences of RPNs as a training project for Canadian students. This information will inform institution-level policy and practices to address the systemic challenges that have plagued the sector for years, offer solutions from invaluable front-line RPNs, and will be accessible to researchers and public audiences, globally

Interpretability of machine learning models that predict cognitive impairment from human speech and language

Machine learning has great potential in detecting cognitive, mental and functional health disorders from speech, as acoustic properties of speech and corresponding patterns in language are modified by a variety of health-related effects. Specifically, neural language models, have recently demonstrated impressive abilities in tasks involving linguistic knowledge. Their success in language understanding and classification tasks could be attributed to their effective representations of linguistic knowledge. However, the increasing complexity of the state-of-the-art models make them behave in a black box manner when the models are not easily interpretable. The successful adoption of machine learning models in healthcare applications relies heavily on how well decision makers are able to understand and trust their functionality. Only if decision makers have a clear understanding of the model behavior, can they diagnose errors and potential biases in these models, and decide when and how much to rely on them. As such, it is important to create techniques for explaining black box models in a human interpretable manner.

Searching for In-Situ Resources on the Moon

As lunar exploration is ramping up, resources present on the Moon are valuable and the distribution of these resources will directly affect future human surface operations. Water ice and other volatiles will be a key factor in surface operations so small rover prospecting missions to identify these resources are essential. In order for such focused rover missions to be successful, a well-developed concept of operations (ConOps) is required. Developing these ConOps through analogue missions as well as testing rover and instrument capabilities through lunar simulations will ensure these prospecting missions are prepared for efficient surface operations and can achieve their science objectives within limited timeframes.

Improved Model Compression Techniques and Processes for Large Scale Pretrained Language Models

Over the past few years, the abilities and performances of deep-learning natural language processing (NLP) have evolved dramatically. A main reason for those improvements is the scaling of the number of parameters in models to tens or hundreds of billions. However, it also becomes impractical to deploy those models in production as the computation cost becomes prohibitive for the vast majority of applications. The project aims to develop a structured compression process incorporating the state-of-the-art techniques (and potentially new ones) allowing quick move from large-scale models to smaller, faster production-ready ones. Additionally, we will investigate the impacts and traits of the devices in industry to further optimize the compression techniques. The ultimate goal is to result in the development of general-purpose compression techniques which can be applied to the whole range of Turing language models. This project can benefit several NLP-related Microsoft services on various devices.