Projets novateurs réalisés

Assessment and modelling of salmon streams in western Newfoundland

The approach to assessing Atlantic salmon in Newfoundland is based on patterns of salmon production and habitat that were established for populations in rivers much further south. The physical and biological characteristics of watersheds and rivers in Newfoundland are different from the southern rivers and the bases for conservation decisions are likely inappropriate. Establishing relationships between stream physical and biological characteristics that are unique to Newfoundland is prudent for better management of Newfoundland salmon populations.

Advancing Gender Equality for Women in Venture Capital – Part 2

Women are vastly underrepresented in venture capital (VC) firms. There is also a lack of women entering the field of venture capital due to systemic barriers and a lack of momentum to shift existing barriers for women to enter the venture capital world. This research aims to understand the current status of women in VC firms, identify structural barriers facing them, and promote gender parity in VC firms and the investing community. The methods include a literature review, in-depth interviews with VC capitalists or senior leaders from VC firms, as well as the creation of an action plan for the partner organizations to drive changes in the investing community.

Cross-Modal Recipe Retrieval

The goal of cross-modal recipe retrieval is to design systems that are able to find a digital recipe, given the user’s image of the food, or find its image, given its ingredients or cooking instructions. For such a cross- modal retrieval task, a common image-text representation space is needed to embed the semantic information of each modality along with the cross-modal mutual information. With the advent of large- scale datasets, such as Recipe1M, the scalability-accuracy tradeoff of the cross-modal embedding methods has increasingly gained more attention in the last few years. The main goal of this project is to use (and improve) the SOTA cross-modal embedding methods to efficiently retrieve a recipe in a large dataset of recipes with low latency and computational demand, and recommend similar recipes based on the queried recipe.

Lifelong learning for service robots

Traditionally, software for home products could not be changed once they were shipped. Furthermore, they could not run complex machine learning (ML) models because their computing and storage capacities are limited due to budget constraints. Recent advancements of cloud infrastructure, however, may allow such products to collect a large amount of data and continuously update the software system throughout their lifetime. Lifelong learning can potentially improve the user experience by making the software system more intelligent and adaptive to specific user preferences. On the other hand, we need to solve problems such as incremental learning, data representation, and catastrophic forgetting. There may also be security and privacy issues associated with processing user data on the cloud.

Multimodal Procedure Understanding

Procedural content (text and video) is abundant on the internet, and is regularly used in our daily lives, e.g., when we follow a cooking recipe to make a dish, or watch an instructional furniture assembly video. Automatically understanding such content allows for the development of various types of AI assistants, including those that can provide answers to our questions (e.g., asking a cooking assistant how much milk we needed in the recipe), and those that can guide users follow through a procedure (e.g., if the user forgets an important step). This project focuses on various aspects of automatically understanding procedural content.

Rapid photocatalytic determination of soil organic carbon content: development and validation of protocols

Global warming is a well-known global phenomenon that has been in the headlines for the past few decades. Greenhouse gases (Carbon Dioxide, Methane, Carbon Monoxide, etc.) are the major contributors to global warming and several governments have been working towards limiting such emissions by introducing emission guidelines. Carbon Dioxide is the gas of interest in this research project as recent studies show that Canadian agricultural soils could remove 11.9 million tonnes of CO2 from the atmosphere annually. The intern will be conducting research to investigate analysis techniques for organic carbon in soil through optimizing previous common analysis methods and under multiple real-life conditions to ensure the reliability of the project’s outcome that would enable the readers to explore a new approach for the analysis of organic carbon in the soil, carbon credits, and providing a green sustainable future for the coming generations.

University of Guelph and IndigenousTech.ai Mentorship Program for Indigenous Youth

IndigenousTech.ai Corporation (IndigenousTech.ai) is is one of the corporations inside of the Forrest Green Group of Companies. All Forrest Green corporations perform web development, solution development, project management and professional services in the technology industry. IndigenousTech.ai has become certified in Ontario and to our knowledge is Canada’s first Indigenous-owned consumer credit reporting agency. IndigenousTech.ai’s mission is to train and hire Indigenous youth on-reserve to increase self-sufficiency, own-source revenue and economic development. To this end, IndigenousTech.ai conducts summer mentorship programs for Indigenous youth focused on financial, digital and accounting literacy. This summer mentorship program is designed to achieve IndigenousTech’s mission described above. Success of the mentorship program requires the contribution from knowledgeable personnel in the fields of technology, business, finance and accounting. This collaboration with the University of Guelph will enable IndigenousTech.ai to conduct a successful Indigenous mentorship program and grow further.

Seleste Western

Seleste is developing an app to integrate with our smart glasses hardware. For this project the intern will help us develop this app which includes two major components. The focus will mainly be the React Native app where the intern will add features to our app and setup our in-house volunteer network for users to call and adding edge ML to our app. Our interface is going to be unique from most since our users are visually impaired and it is incredibly important to create an interface that is simple, accessible and easy-to-use with a screen reader. The second part of the app is adding the ability for guides to remotely take a photo from the glasses. This involves computer networking and sending data from the glasses to the guide using both Wi-Fi and LTE. Both parts of this project will be a core part of Seleste’s IP and value proposition. Successful completion of both parts of this project will allow us to start shipping out our initial pre-order of glasses and run a successful pilot.

Development and testing of a business development strategy for the world’s first and only ubiquitous edge voice AI platform

Picovoice is the first and only ubiquitous edge voice AI platform. The world’s first production-grade local Speech-to-Text technology has been recently added to its portfolio. Currently, there is no direct competition on the market, but alternatives. The intern will be responsible for developing and testing business development and go-to-market strategy with vertical-specific value propositions and communications.
Voice AI is a big market mainly dominated by big players such as Amazon, Google, Microsoft and IBM. Those offerings are all cloud-based. Picovoice differentiates itself with its easy-to-train developer console and its superior speech recognition technology that achieves higher accuracy than the cloud-based alternatives with the benefits of edge computing: up to 100x cost-effectiveness, privacy, and zero-latency.

Finding the niche for on-device, private and high volume use cases and the correct positioning of Picovoice technology within this crowded market will contribute to the success of Picovoice. The project will include crafting the strategy and also execution to test whether the strategy works and iterate accordingly when needed.

Integrating DC Fast Charging into the Unico Power Cerebro Energy management System

Electric vehicle adoption is well past the tipping point in Canada and the world, and although the demand for electric vehicles is no longer a question, the supply chain for electric vehicles is still very much in doubt. One key element is the ability to supply electricity to support the charging of the vehicles, and Unico Power is focused on providing a demand-managed approach to charging that monitors building and grid load to ensure that electric vehicle charging does not use unnecessarily high power levels. Unico Power has successfully developed and delivered a commercial energy management system that supports demand-managed level 2 electric vehicle charging to private owners.

However, projected rapid growth in DC Fast Charging (sometimes referred to as Level 3) can have a much more significant impact on sustainable growth since each DC fast charger can use 250 kW (same amount of peak power as 70 three-bedroom homes). The challenge is to ensure that power demand is managed to support charging more cars without spikes in grid peak demand.

AIDOX – Document Verification System

The existing document understanding systems use machine learning methods, natural language understanding and text analysis, to validate structured trade contracts for language and economic term correctness. The system is now being expanded to allow general document understanding across a wide variety of financial documents, beginning with a focus on customer provided reference material. The proposed solution will extract the key data elements from the documents and validate this data against the internal source of record. This provides several new challenges in document classification, visual document understanding and entity extraction. For structured documents a template-based solution is to be employed to extract key elements. For semi/un-structured documents, a multi modal framework is used incorporating text, layout and image information to extract key elements. Current state of the art solutions (BERT, LayoutLM) still fall short of human abilities, however by properly constraining our problem space we strive to achieve better results.

Mobile App Development Internship

Sliike is developing an App that is unique and in the market space. The App is also tailored for the English speaking and French customers and clients in Canada. We are currently at the advance stage of the Mobile App design. At this stage, we need more hands to support our Internal Mobile App Developer and to join the Project Team to bring the App to life.