Innovative Projects Realized

FABJRP, Towards a Fully Automated Bilingual Job Recommendation Platform

Recruitment of future employees is an essential activity in any organization, yet it is tedious and error-prone. Substantial effort is spent in rote tasks like finding candidates matching a particular job offer, contacting them, scheduling an interview and performing the actual interview, while the more interesting tasks like making a final decision on who to hire are more exciting, yet risky. This project aims to explore the extent to which a recruitment platform could fully automated the hiring cycle, by leveraging AI technologies. In particular, we will automatically identify a shortlist of job candidates for a given offer, and make a final recommendation based on the interview responses. The interviews will be performed automatically by an AI chatbot who can engage job candidates and react to changes in emotions (e.g., stress or frustration). This project will help our industrial partner explore the limits of today’s AI and software technologies for AI-based recruitment.

Ciena OPn Innovation WP 1.1.6 – High Speed Low Power Transceiver

The intent of this project is to address the high-speed electronic portion of a silicon photonic transceiver solution that will explore new and innovative metro reach terabit optical modems. In total there are five projects that combine to create the solution. These five project areas are silicon photonic design, high-speed electronic design, modelling, packaging and test.
The throughput of Ciena’s next generation optical modems is approaching a Terabit per second, transporting data within the chip, across different dies within the same package, and between different modules on the card is becoming one of the limiting bottlenecks to our systems. To overcome this limit a 100Gb/s capable SERDES is required. Indeed, the next frontier that needs to be surpassed is a design of SERDES link in the most economical way in terms of power and real-estate. Our world-class analog/mixed-signal design team at Ciena has expertise covering high-speed data converters, multiplexors, de-multiplexors PLLs and CDRs. In this project the collaboration will be between Ciena and Prof. Mohamad Sawan, from Polytechnique Montréal. This research project will focus on designing, implementing and testing high-speed data transmitters.

Accelerating discovery through high-throughput experimentation and machine learning – Year two

Canonical methods of molecular discovery and reaction optimization rely on “trial-and-error” approaches and slow experimentation with low discovery rates. By harnessing high-throughput experimentation (HTE) with machine learning (ML) methods, artificial intelligence (AI) and robotics, we have the potential to dramatically accelerate the discovery and preparation of next generation molecules and materials. We will extract, unify, and transform data from literature into actionable intelligence, and generate a robust workflow for the automated synthesis of catalysts and resins at NOVA Chemicals. Through ML models, we will leverage newly-generated data to guide experiments and simulations, enabling rapid molecule development, and culminate in the inverse design of molecules and materials targeting function rather than a particular molecular structure. By combining the expertise, software, and hardware tools of the Hein Lab with the instrumentation and extensive database at NOVA Chemicals, we will create a closed-loop, self-driving laboratory that will (i) be capable of implementing a diverse range of chemical workflows and (ii) create datasets that will be leveraged by AI, allowing users to navigate complex structure-function relationships and experimental landscapes.

Accelerating discovery through high-throughput experimentation and machine learning

Canonical methods of molecular discovery and reaction optimization rely on “trial-and-error” approaches and slow experimentation with low discovery rates. By harnessing high-throughput experimentation (HTE) with machine learning (ML) methods, artificial intelligence (AI) and robotics, we have the potential to dramatically accelerate the discovery and preparation of next generation molecules and materials. We will extract, unify, and transform data from literature into actionable intelligence, and generate a robust workflow for the automated synthesis of catalysts and resins at NOVA Chemicals. Through ML models, we will leverage newly-generated data to guide experiments and simulations, enabling rapid molecule development, and culminate in the inverse design of molecules and materials targeting function rather than a particular molecular structure. By combining the expertise, software, and hardware tools of the Hein Lab with the instrumentation and extensive database at NOVA Chemicals, we will create a closed-loop, self-driving laboratory that will (i) be capable of implementing a diverse range of chemical workflows and (ii) create datasets that will be leveraged by AI, allowing users to navigate complex structure-function relationships and experimental landscapes.

Use of a deep passive source extremely low frequency (ELF) conductivity mapping system to improve the definition of ore bodies at depth – Applications to Bathurst, NB

The Bathurst Mining Camp, located in northern New Brunswick, is one of Canada’s oldest mining districts. Most of the 46 known deposits were discovered in the 1950s using a combination of geological and geophysical methods. However, renewed exploration efforts over the past 15 years have not been as successful as one would expect for the level of expenditure the camp has gone through.
Aurora Geosciences Limited (AGL) is a leading-edge service provider in the application of geology and geophysics for mineral exploration. Over the past 5 years they have been using a passive source electromagnetic system (called ELF) that has the advantage of not requiring any active sources of energy (e.g generators) nor cables to be laid out in the ground. Another advantage of this system is that it can generate deep images of the subsurface up to 1-1.2 km depth. The goal of this project is to use Aurora’s ELF system over a series of known deposits in Bathurst and then use these data to produce 3D models. These models will help further exploration efforts in the area.

Disparition et formes de vie: étude sur le geste de disparaître et ses possibles dans la littérature et les arts actuels

Ce projet de recherche s’inscrit dans le cadre de ma cotutelle de thèse. Intitulée «Disparaître autrement. La disparition comme geste, et son rapport aux formes de vie dans la littérature et les arts actuels», ma thèse s’intéresse à un corpus tant littéraire qu’artistique, et s’appuie sur un édifice théorique principalement philosophique. Je cherche à analyser la représentation du «disparaître» au regard des formes de vie et des reconfigurations possibles du rapport entre l’individu et le social, dans des œuvres littéraires et artistiques (arts visuels, médiatiques et de performance) produites en Europe occidentale et en Amérique du Nord depuis 1990. Le projet de recherche que je présente aujourd’hui vise principalement à développer cette structure théorique liant littérature, philosophie, politique et représentations du réel, qui me permettra de problématiser les œuvres de mon corpus principal de thèse.

Investigating the effect of novel nutritional compounds on skeletal muscle protein synthesis and growth in vivo

Resistance exercise training combined with adequate post-exercise protein ingestion is known to increase muscle mass when completed repeatedly over a prolonged period of time (>6 weeks). Due to this knowledge much research has been conducted to identify the best protein supplements which allow for the greatest muscle growth with training. This project will test if a novel protein supplement, which also includes natural compounds believed to stimulate muscle stem cells, can increase muscle size when combined with resistance exercise training when compared to post-exercise branched chain amino acid or carbohydrate ingestion. The results of this project will then inform the industry collaborator, Iovate, on the effectiveness of this supplement and how best to market it.

Retrospective molecular subtyping of pediatric medulloblastomas and the evaluation of BTIC gene signature in tumors with poor prognosis

Medulloblastoma is the most common brain tumor in children. It is treated with a combination

of surgical resection, chemotherapy and radiation. Radiation to a child’s brain can have

harmful side effects that may have implications in later development. We intend to use

molecular gene expression to classify archived tumors into 4 subgroups with associated low

and high risk. Along with this, we will analyze the expression of genes associated with a

highly resistant subpopulation of cells called brain tumor initiating cells (BTICs). These BTICs

may be responsible for cancer recurrence and be driving factors in high-risk cases. Tumor

RNA will be isolated from paraffin blocks and analyzed for expression of 43 different genes

using NanoString nCounter technology. Molecular gene expression and clinical outcome will

be correlated to better understand trends in this disease. We hope to avoid irradiating mild

cases unnecessarily while ensuring aggressive treatment for poor prognosis

medulloblastomas.

Geographic mapping for small-diameter gas pipelines in a city

Geographic location of a pipeline is important information for pipeline maintenance and fault detection. Usually, the geographic location of a pipeline on the ground can be measured directly with global positioning system (GPS) technology, but it is much difficult to determine the geographic position of an inaccessible underground pipeline in a city. In this research, a new geographic mapping methodology is proposed for small-diameter gas pipelines in a city. A pipeline mapping micro robot equipped with a micro electro mechanical system (MEMS) based inertial measurement unit (IMU) and odometers is developed. The technique of motion identification and measurement is proposed based on the odometers. A new estimation approach is proposed to reconstruct robot path based on the features of a pipeline, the bending angles and the identified motion modes. The geographic location of a pipeline is obtained by the path optimization algorithm. The proposed methodology can help the partner organization obtain the 3-dimensional (3D) map of old gas pipelines in a city, facilitate the maintenance and leak position detection, so as to reduce troubles in the pipeline and avoid economic losses.

Improving the Accuracy of Data Loss Prevention Systems

Scotiabank employs teams of cybersecurity specialists across its global operations and partners with a variety of external organizations to prevent and investigate any electronic attempts to gain access to the Bank’s data. At the same time, employees are continuously educated and expected to look for warning signs and efforts to infiltrate that data as well. Currently, Scotiabank’s Data Loss Prevention (DLP) systems have a high false positive rate in identifying data breaches and cyber-attacks, which require significant manual intervention. The project will use machine learning algorithms, data mining principles, and cybersecurity threat modelling to improve the accuracy of Scotiabank’s DLP systems by reducing the false positive rate. We will also develop automated reporting systems and reduce the need for manual verification of data loss events. This project will benefit the partner by improving upon the accuracy and automation of their DLP systems, as well as benefiting Canadian consumers who rely on Scotiabank to keep their personal data safe.

Tracing CO2 behavior using HGS and PEST in the shallow aquifer of the K-COSEM research site, Korea

To reduce greenhouse gases, Carbon Capture and Storage (CCS) can be an option. This method is being used all around of the world for stopping global warming. The goal is to remove an excess CO2 in the air. My research uses modeling methods, such as HydroGeoSphere (HGS) and Parameter ESTimation (PEST), which will evaluate how CO2 moves when CO2 is artificially injected in groundwater. Also, I develop the way to find out CO2 leak location and time when CO2 is leaked. Successful completion of this study will make it easier to identify CO2 leakage location and time when CO2 emissions are detected. It could be applied not only to CO2 but also to other sources of contamination of groundwater.

Development of fast photocrosslinking bioinks for visible light-based stereolithography 3D bioprinting

The research projects aims to develop the fast photocrosslinking bioinks for visible light-based stereolithography 3D bioprinting.
In order to prepare bioinks for 3D bioprinting, alginate was modified with tyramine which is phenol group and gelatin was modified with glycidyl methacrylate. Compound with alginate-tyramine derivatives(Alg/Tyr) and gelatin-methacrylate(GelMA) will improve its 3D printability, mechanical properties and cell interaction.
The requirments of bioinks for 3D bioprinting are fast gelation time, excellent cell proliferation and mechanical properties. The gelation time of bioinks will be promoted with tyramines of Alg/Tyr, methacrylates of GelMA and tyrosines of gelatin. In order to supplement the limited cell adhesion of alginate, gelatin will play good cell interactions. With its photo-crosslinking between several phenol and methacrylate moieties, the mechanical properties of Alg/Tyr and GelMA compounds will be enhanced.
Consequently, this project will be successfully accomplished by Professor Keekyoung Kim’s diverse research experience and the ability to conduct advanced research on 3D bioprinting.