Innovative Projects Realized

Self-Improving Agents with test-time compute for FDA

Current artificial intelligence (AI) agents deployed in enterprise settings make static, one-shot decisions and cannot adapt as context changes or learn from the outcomes of past actions. This research addresses two limitations of enterprise AI agents: (i) the inability to dynamically adjust reasoning effort based on the uncertainty and stakes, and (ii) the inability to improve their decision-making capabilities over time by learning from observations. Both result in agents that are brittle under distributional shift, poorly calibrated in their confidence, and unable to close the loop between prediction and consequence.
ServiceNow aims to build Foresight-Driven Agents (FDA): enterprise AI agents that maintain a dynamic world model of an organization, reason over evolving context, and forecast the consequences of decisions. The specific challenges this project addresses are: (a) how to allocate additional test-time reasoning to highuncertainty, high-stakes decisions, and (b) how to design self-improvement mechanisms that allow the agent to update its world model, recalibrate its confidence, and refine its decision strategies based on delayed and noisy outcome signals.
ServiceNow develops a platform to manage and automate large-scale enterprise processes across diverse industries. In 2020, ServiceNow acquired ElementAI to boost its standing in AI research and the Canadian AI scene, leading to enhanced AI-powered products and platform features. ServiceNow AI Research has made significant contributions to the field of foundation models and has a strong presence in developing AI agents for different domains.
This project will advance ServiceNow’s capacity to deploy agents that more reliable, well-calibrated decisions in dynamic environments. By enabling agents to allocate reasoning effort where it matters most and to improve from experience, the project directly addresses operational efficiency. The resulting research contributions, including open-source benchmarks and publications at top venues, will strengthen the Canadian AI research ecosystem and ServiceNow’s position as a leader in enterprise AI.

ARCHAEOLOGY EXPERT SYSTEM

(1) Overview of the Research Problem
Despite rapid advances in Large Language Models (LLMs), their deployment in professional knowledge-intensive sectors remains constrained by hallucinations and data gravity. LLMs frequently generate plausible but unsupported outputs when applied to specialized corpora, while organizations hold vast quantities of “dark data” embedded in unstructured PDF archives. In archaeology and heritage management, this challenge is amplified by decades of technical reports, stratigraphic records, regulatory interpretations, geospatial metadata, and catalogues comprising several million artifacts dispersed across sites. Standard Retrieval-Augmented Generation (RAG) systems relying solely on vector databases often struggle with structured, relational, chronological, or geospatially constrained queries [1]. This research proposes a hybrid architectural approach integrating large-scale PDF vectorization, dual (vector and relational) indexing, geospatial metadata integration, and parameter-efficient model alignment to ensure strict factual rounding and traceability.

(2) Partner activities and Challenges
Archéoconsultant operates in professional archaeology and territorial heritage management while also developing and commercializing digital solutions for the archaeological sector in Canada and internationally. The organization manages a substantial archive of domain-specific documentation essential for regulatory guidance, mitigation planning, and environmental impact assessments. The partner seeks to bridge the “trust gap” in AI adoption by developing a domain-constrained knowledge infrastructure capable of delivering verifiable, source-grounded responses while enabling spatially aware and chronologically precise analysis.

(3) Anticipated Social and Economic Benefits
The project converts days of manual research into near-instantaneous, high-precision queries, generating significant economic value by accelerating development timelines and reducing compliance risks for large-scale infrastructure. By exploiting structured data, the system strengthens governance through more transparent, traceable, and evidence-based regulatory decision-making. Ultimately, this scalable framework fosters social impact by enabling proactive cultural heritage protection and positioning the Canadian heritage sector as a leader in responsible, modern AI adoption.

Community-informed interventions to reduce air pollutant emissions in near-port neighbourhoods

Strathcona, a neighbourhood in Vancouver, experiences higher air pollution than many other parts of the region as it sits next to major freight routes and industrial areas. Studies have found persistently high levels of nitrogen dioxide (NO2), a pollutant linked to traffic and fuel combustion that can harm human health. While the problem is well documented, it is less clear which actions could reduce pollution most effectively in the next few years. Pollution in Strathcona comes from several sources, including heavy-duty trucks, trains, ships, and nearby industry. Responsibility for these sources is spread across multiple levels of government and other decision-makers, making it challenging to design and implement solutions. This project works in partnership with the Strathcona Residents Association to identify practical, near-term actions that could lower pollution. Through analyzing documents, emissions modelling, and stakeholder input, the project will prioritize strategies that are both effective and realistic, helping the community focus its advocacy on measures most likely to improve local air quality and health.

Access to Health Care for Uninsured Migrants – Strategies for Community Advocacy

This project examines how uninsured migrant parents of babies born in Canada, and their supporters, navigate barriers to registering newborns for provincial health cards. “Uninsured migrants” refers to people who are ineligible for public insurance and lack private coverage or have insufficient coverage. Research shows that hospitals may refuse to register newborns when the mother is uninsured, and many parents avoid seeking care due to fears about exposing their immigration status (Bernhard, Young, & Goldring, 2025; Cheff & Bates, 2018). As a result, some newborns miss essential early care, and families face high out-of-pocket costs or delay care (Gagnon, 2026).
The Immigrant Workers Centre (IWC), a Montreal-based organization supporting workers with precarious immigration status, has identified growing challenges related to pregnancy, post-natal care, and newborn health card registration. In partnership with the McGill School of Social Work, this project will conduct a literature and policy review (Quebec and Ontario), interviews with 6 uninsured parents and 14 service providers, and develop an adaptable inventory of effective advocacy strategies to support cross-provincial learning and action.

Solis Medical Clinic – Strengthening Service-Line Positioning and Digital Marketing for Multidisciplinary Care

This project will help Solis Medical Clinic improve how it communicates its services and reaches new patients by strengthening its marketing systems, service-line messaging, and patient-education materials. The proposal notes that many prospective patients are unaware of the clinic’s full range of treatments and that Solis currently relies on “fragmented communication channels, limited analytics, and inconsistent messaging across service categories.” The intern will conduct market research, refine service descriptions, improve digital marketing pathways, and build educational tools that help patients better understand their care options. These improvements will support clinic growth, increase patient awareness, and enhance access to multidisciplinary pain and wellness services.

Smart City Modeling with Agent-Based and Discrete-Event Simulation

This project will develop and compare two simulation approaches for a Hamburg smart-city scenario and package the workflow so it can be reused for similar modelling in Ottawa. I will build (1) an agent-based model using the MARS framework (C#) with layered/GIS-style representations where agents move and interact based on their surroundings, and (2) a comparable discrete-event/DEVS model using Cadmium/DEVSsim for the same scenario.

The goal is to understand how modelling assumptions and “layer” design affect outputs and how each approach supports decision-making for urban mobility and infrastructure planning. I will run aligned experiments in both models and evaluate results using consistent metrics such as flow/throughput, congestion proxies, travel-time distributions, and spatial density patterns. Python will be used for data processing, visualization, and reporting.

Deliverables: two working models (MARS + DEVS), a set of experiment scripts, a short comparative report, and a brief “setup/adaptation guide” to help apply the same workflow to Ottawa.

Unlock the power of open data in commercial insurance (Desjardins)

In commercial insurance, businesses are constantly changing and keeping up is no small task. As their activities grow or evolve, their exposure to risk and insurance needs change. Outdated or incomplete information about the customer can lead to poor customer experience and unfair pricing. While regularly reaching out to clients could help keep their information up to date, this approach would be costly and potentially disruptive to their experience.
By using external open data sources that are frequently updated by different organizations, municipalities and government bodies, we aim to find ways to validate our internal information about business owners, and use it in our processes and even find trends and patterns in those data that can translate to growth opportunities within the insurance market.

Westmount Dental Lab Innovation

The project aims to improve the digital denture production workflow at Westmount Dental Lab. The intern will analyze current processes, identify inefficiencies, and develop standardized procedures to improve quality.

Reimagining Land-Use Planning: Black Agrarian and Food Futures in Colombia

This project hopes to better understand how land-use planning mechanisms can be reestablished for the benefit of traditional Black lands in the Alto Cauca region of Colombia. Using spatial data derived from community-based research processes, this research hopes to support the development of innovative planning processes aimed to highlight Black land rights in Colombia. This research will support larger efforts of the national government to recognize traditional Afro-Colombian and Indigenous land rights, particularly in promoting their respective political autonomy as well as providing financial support to maintain traditional practices, ways of living, and relationships with the physical environment. Afro-Colombian and Indigenous use of alternative governance strategies, collective land ownership, and traditional ecological knowledge are widely acknowledged for their substantive benefits towards the conservation of forests, ecosystems, and biodiversity. This work will support 18 Black Community Councils, scholars at the Universidad Autónoma de Occidente, and the NGO Funecorobles, all of whom have been collectively working together to better understand how Black land rights in Alto Cauca could be better realized. Research findings will be disseminated with local planners through workshops aimed towards the final stages of the project, in hope to sensitize planners to the limitations of current statutory planning processes.

Exploring Artificial Intelligence applications and considerations in Patient Health Education

The RBC Patient and Family Learning Space (PFLS) at CAMH provides free mental health, substance use, and recovery information to patients, families, caregivers, and the public. PFLS offers workshops, webinars, easy-to-understand health information, and help with navigating the mental health system. It operates both as a physical space and an online learning hub, working with CAMH teams and community partners to improve access to information and reduce barriers related to health literacy and social factors.

As PFLS grows, there is a need to improve how educational materials are created, updated, and shared. Right now, many processes—such as making information easy to understand, translating materials into different languages, and adapting content into formats like videos or audio—are done manually. This takes significant time and staff effort, and makes it harder to keep materials current and accessible to diverse communities.

This summer project will support PFLS to develop and test new, improved ways to create and share mental health education materials, including the responsible use of artificial intelligence (AI) tools. AI can help translate content, adapt materials into plain language, and create different formats more efficiently. However, careful review is needed to ensure information remains accurate, culturally appropriate, and trustworthy.

The project will be carried out by two interns, providing them with valuable hands-on experience in public health, mental health education, knowledge translation, and emerging digital tools. Under supervision, interns will help review existing materials, explore AI tools, and contribute to developing clear guidelines and workflows. This project will strengthen PFLS’s ability to provide high-quality, accessible education while also building workforce capacity and supporting the next generation of public health professionals.

Menopause Symptoms and Cogniton: Real-world evidence from reproductive health tracking applications

This project at the Centre for Addiction and Mental Health addresses a defined evidence gap in women’s brain health: despite women comprising roughly two-thirds of Alzheimer’s disease cases, there is limited scalable, person-centred data on how specific menopause symptom patterns and menopausal hormone therapy (MHT) exposures relate to cognitive outcomes. In partnership with two major reproductive health tracker apps, Clue and Natural Cycles, the project will use large-scale app-based data to identify data-driven menopause subtypes based on symptom co-occurrence using clustering methods, and model how these symptom clusters, symptom frequency/severity, and MHT type and route (e.g., estradiol vs. progesterone; pill vs. patch) predict cognitive complaints such as brain fog, forgetfulness, and difficulty concentrating. Findings will directly inform feature-engineering and risk-modeling strategies for a larger lab initiative to develop a female-specific Alzheimer’s disease risk calculator drawing on 350,000+ participants across multiple aging databases. The expected benefit to CAMH is the development of reproducible analytic pipelines, empirically derived menopause subtypes, and predictive models that strengthen its capacity to build targeted prevention tools, improve early risk identification in midlife women, and enhance Canada’s leadership in data-driven women’s health research at the intersection of science and industry.

Integrated Renewable Energy Sharing and Decentralized 5G Infrastructure for Sustainable Rural Development

Reliable electricity and internet access are essential for basic services (healthcare, education, emergency response) and economic opportunity, yet many rural and off-grid communities, including some Indigenous areas, still lack both. Thus, this research project investigates how rural and off-grid communities can access more affordable and reliable electricity and broadband by planning energy and connectivity together rather than separately. It proposes a peer-to-peer (P2P) energy-sharing network in which households equipped with solar PV, small wind turbines, and battery energy storage (BESS) can share electricity locally. At the same time, the same local clean energy also supports decentralized telecom service through 5G small cells. A series of analyses will be conducted to show that integrating renewables, storage, and decentralized telecom can reduce energy costs, increase self-sufficiency and resilience, cut reliance on fossil fuels, and support social and economic development (e.g., education and local businesses). The analysis will be developed to examine the roles of equipment prices and battery capacity in economic viability.