Projets novateurs réalisés

AI-driven Spatiotemporal Analysis of Human–Nature Interaction and Cultural Ecosystem Services at the Global Scale

This project aims to use artificial intelligence to study how people interact with nature in cities and natural areas around the world. By analyzing millions of social media photos and captions, the research will identify different types of human activities, emotions, and cultural values connected to nature. The project combines computer vision, natural language processing, and geographic analysis to understand global patterns of environmental engagement. It will benefit the University of British Columbia and the Massachusetts Institute of Technology by strengthening their collaboration, sharing advanced research methods, and producing a global dataset and visualization tools that can guide sustainable landscape design, environmental planning, and public policy.

Supporting Review of Topic-Based Code Changes

La revue de code est essentielle pour assurer la qualité logicielle, et une revue efficace nécessite de comprendre à la fois la modification individuelle et le contexte plus large dans lequel elle s’inscrit. Dans de nombreux projets, y compris ceux utilisant Gerrit, les développeurs regroupent les modifications liées en un « topic » qui représente une fonctionnalité complète répartie sur plusieurs pull requests.

Lorsque ces modifications liées sont examinées de manière isolée, les outils d’IA produisent souvent des commentaires incorrects ou répétitifs, car ils ne peuvent pas voir comment les différentes parties s’articulent. Par exemple, une fonctionnalité de mode sombre peut être répartie sur plusieurs changements concernant les couleurs, les composants d’interface et la logique, et elle ne prend tout son sens que lorsqu’elle est vue comme un topic.

Ce projet vise à développer des modèles d’IA pour la revue de code qui soient « topic-aware », capables d’analyser des modifications liées comme un ensemble cohérent. Nous créerons TopicReviewBench, un jeu de données composé de topics réels, et nous évaluerons plusieurs grands modèles de langage, tels que GPT-4, Claude et Llama, en testant différentes manières de leur présenter le contexte inter-modifications.

Procédés de microfabrication pour la polarisation de guides d’ondes dans les isolateurs

AEPONYX Inc. a développé une technologie innovante de guides d’ondes dans les isolateurs, révolutionnant l’intégration des isolateurs aux circuits photoniques et quantiques haute performance. L’intérêt du marché pour cette technologie est manifeste, mais la capacité de production demeure limitée en raison du caractère expérimental du procédé actuel. Le projet général vise à concevoir un procédé manufacturable et l’infrastructure associée afin de répondre aux besoins du marché.
La stratégie repose sur la technologie de reconstitution de tranches par moulage (FOWLP), qui permettra une fabrication industrielle robuste et à grande échelle. Cette approche consiste à regrouper des dispositifs discrets sous forme de tranches pour leur appliquer des étapes de microfabrication et d’intégration dans des équipements automatisés de fonderie. En combinant les guides d’ondes isolants et la reconstitution FOWLP, nous pourrons adresser des marchés à fort volume avec une technologie entièrement « made in Canada ».

L2M – TruckBud Commercialization Strategy

The project will focus on applying structured market research methods to better understand the commercial potential of my venture idea. I will interview potential customers, analyze market insights, test assumptions, and evaluate different business model options. This work will help me clarify the problem I am solving, identify real customer needs, and determine whether the opportunity is strong enough to move forward. The expected benefit to Lab2Market is that this research directly supports the goals of the program by helping a participant develop evidence-based insights, improve the quality of their commercialization work, and increase the chances of building a viable, scalable venture.

CLIQMedia – AI-Enabled Digital Marketing Insights Framework for Predictive Campaign Innovation

CLIQMedia is a Vancouver-based digital marketing agency specializing in web design, development, branding, and advertising for clients in construction, real estate, design, and dental industries. As the company continues to scale, it faces an innovation challenge around how to generate deeper, data-driven insights that meaningfully improve campaign outcomes in industries that are rapidly digitizing but historically slow to adopt advanced marketing technologies. Currently, the agency relies on manual research, traditional competitive analysis, and platform-native analytics, which provide only surface-level insights and limit the ability to anticipate market trends or design highly targeted campaigns. The improvement priority is to build a more sophisticated, predictive, and automated system for gathering, analyzing, and applying digital marketing intelligence—something that goes far beyond day-to-day campaign execution. This project proposes to explore next-generation marketing technologies, including AI-enabled trend analysis, competitor intelligence mapping, and automated insights dashboards, to help CLIQMedia redesign how strategic decisions are made for clients. Solving this problem requires expertise in marketing strategy, digital research, analytics interpretation, and the ability to translate complex findings into implementable recommendations. The intern will support the exploration and design of a new insights framework that modernizes CLIQMedia’s processes, enhances client ROI, and creates a scalable model for future digital marketing innovation across the agency.

L2M Voltage Meter

The Voltage meter is a first of its kind device created with the intention of bridging the gap between the symptomatology we understand from our frail patients’ (those who score 0-1 on the Frailty Index) experience and a numerical way to quantify rehabilitative progress or decline in the clinical environment. We know that there is a relationship between changes in voltage and an increased level of frailty as we age, but until now we have not been able to create a device to monitor this change in health status. From my first hand experience as a Preliminary Neurosurgeon, I have seen this gap in both my own clinical practice and the research settings around me. The goal of the voltage meter is to quantify numerically the patients’ specific nerve amplitude to better understand their clinical status, rehabilitation developments or clinical decline. The hope is to implement this into both care settings and rehabilitative environments as an ‘easy to use tool’ to assess the patient’s level of frailty, functional capacity and specific voltage changes over time.

L2M – WindChat Intelligence: Conversational AI for Wind & Hydrogen Site Screening

The wind energy industry currently relies on a fragmented workflow involving multiple disconnected and expensive software tools for site assessment, energy calculations, and reporting. This inefficiency limits the number of sites developers can evaluate and creates a high barrier to entry for smaller developers and community projects.

This project, titled WindChat, focuses on validating a conversational AI tool designed to streamline preliminary screening for wind farms and green hydrogen projects. WindChat aims to allow users to describe a potential site in natural language and receive transparent, auditable estimates of wind energy and hydrogen production, along with simple techno-economic indicators.

Through the Lab2Market Validate program, the intern will bridge the gap between academic research in wind resource assessment and commercial viability. The primary objective is to conduct rigorous customer discovery to determine whether wind professionals trust an AI interface for decision-making and under what conditions. Key activities include building a professional demonstration environment using real Canadian wind site scenarios, developing a technical documentation package to ensure transparency and trust in the tool’s calculations, and testing various pricing models with potential customers.

The project involves conducting customer interviews to gather evidence on willingness-to-pay and feature priorities. By the end of the internship, the intern will deliver a validated Business Model Canvas, a functional demo platform, and a go-to-market roadmap. This initiative supports Canada’s transition to a net-zero economy by potentially lowering the cost and complexity of deploying renewable energy infrastructure, particularly for smaller and emerging wind and hydrogen developers.

Graph Neural Networks for Wind Power Modelling

This project aims to develop new machine learning models that can help wind energy companies design wind farms more efficiently and at lower cost. Today, planning a wind farm requires running large, high-resolution computer simulations to understand how wind flows around turbines and how much power a proposed layout can produce. These simulations are accurate but extremely slow and financially expensive, especially when companies need to test a large number of different layouts or new locations to optimize the power production and transmission. Our research will create advanced prediction models that learn directly from existing wind flow simulations and real atmospheric data. By using graph-based neural networks and advanced generative methods, our models can adapt to different turbine layouts and geographic domains, while also estimating uncertainty in the predicted wind and power output. This makes them more flexible and scalable than current tools. For Veer Renewables, the partner organization, this project will produce a new commercial product that can deliver fast, early-stage assessments for new wind farm layouts and new geographic domains, reducing reliance on costly physical simulations.

L2M – RepGen: Towards Automated Reproduction of Deep Learning Bugs Leveraging an Intelligent Agent

This project aims to explore whether RepGen, an academic tool that helps software teams automatically reproduce bugs in deep learning and AI systems, can become a useful product for industry. By interviewing more than 100 developers, companies, and technical teams, the project will identify who needs this tool the most, what features they care about, and how RepGen can fit into real software development workflows. The results will guide the creation of a business model, a product roadmap, and a list of early adopters. This work will benefit the partner organization, Lab2Market, by providing clear, evidence-based insight into the commercial potential of RepGen and supporting their goal of turning university research into practical, market-ready innovations.

L2M – Development of a modular AI voice architecture for automated dental front desk operations

Orbis Assist is developing “Ava,” an intelligent voice assistant designed to function as a fully automated receptionist for dental clinics. This internship project focuses on building the secure software that allows the AI to manage phone calls, schedule appointments, and answer patient questions by directly connecting with the clinic’s internal calendar systems. By automating these critical front-desk tasks, this technology ensures that dental practices can operate 24/7 without being limited by the current shortage of administrative staff, ultimately helping clinics reduce wait times and ensuring more Canadians have timely access to oral healthcare.

Co-adaptation d’une formation sur la gestion du stress pour les producteurs

Le projet vise à co-adapter une formation sur la gestion du stress pour les producteurs agricoles. Ce projet correspond à la première étape d’un plus grand programme, où l’objectif sera d’évaluer l’intérêt et d’identifier les attentes des producteurs laitiers concernant la formation. La perspective sera de concevoir la formation à partir des réponses des producteurs laitiers.

Les versions antérieures de cette formation, qui ne sont pas adaptées pour les producteurs, ont permis aux participants d’augmenter leur proactivité dans la gestion de leur stress. Cela a contribué à diminuer significativement leur symptômes de stress et d’anxiété.

Jimmy Zee’s Distributors – Enhance Sales Efficiency through CRM Optimization and Digital Lead Intelligence

Jimmy Zee’s Distributors Inc. is a Vancouver-based distributor specializing in confectionery and general merchandise that celebrate Canadian culture and identity. The company supplies a broad network of retail partners—from convenience and grocery stores to hardware and tourist outlets offering products that bring joy and pride to consumers across the country. As Jimmy Zee’s continues to grow, the organization faces the challenge of scaling its customer acquisition strategy in a way that is both efficient and data-driven. Currently, lead generation and client outreach are handled through traditional, manual processes that limit the ability to strategically identify high-value prospects, assess conversion potential, and measure the performance of sales initiatives. This project introduces an innovative approach to modernizing these operations by developing and implementing a structured lead generation system that integrates customer relationship management (CRM) tools, data validation methodologies, and performance metrics to guide decision-making. The innovation lies in transforming how Jimmy Zee’s collects, analyzes, and leverages customer data to generate leads more effectively moving from intuition-based outreach to evidence-based strategy. This initiative goes beyond day-to-day sales activities by creating a replicable framework for digitalized lead generation, enabling long-term scalability and market expansion. Solving this challenge requires expertise in marketing analytics, sales process design, and CRM system optimization, along with creative problem-solving and communication skills to translate data insights into actionable business outcomes.