Centre for Applied AI

Artificial intelligence, and in particular recent advances in generative models and large language models, have enabled transformative applications across healthcare, software development, telecommunications, manufacturing, and decision-making in both the public and private sectors. Yet the inner workings of these systems remain largely a black box, and fundamental questions about their trustworthiness are still unresolved. Researchers at the Faculty of Engineering bridge foundational machine learning with real-world engineering applications to tackle these challenges head-on.

Our faculty offers expertise across the entire AI pipeline. We examine the data that fuels large-scale AI systems, investigating issues of data quality, faithfulness, and privacy, and asking how training data should be collected, curated, and protected. We probe the training process itself, developing tools and theories that open the black box and reveal how models learn, what they memorize, and where they can fail. And we rigorously evaluate and benchmark deployed AI models across multiple dimensions, including fairness, robustness, safety, and alignment, to provide principled answers to the question: how trustworthy is this system, and for whom?

Building on these foundations, our researchers study how to responsibly leverage AI across critical engineering domains. This includes AI-assisted software verification and testing, secure and intelligent communications networks, natural language processing, and the development of autonomous agents and robotic systems capable of operating safely in complex real-world environments. Crucially, we do not treat ethics and law as an afterthought: our researchers also examine the ethical, legal, and societal implications of AI technologies, ensuring that the systems we help build are not only technically sound but societally responsible.

Researchers

Bissessar, David
Briand, Lionel
Chen, Boqi
Fattahi, Javad
Hamou-Lhadj, Abdel
Lu, Yiwei
Mao, Yongyi
Millar, Jason
Rahgozar, Arya
Sun, Bo
Yao, Jianping

IES seamlessly integrate renewable energy sources, storage solutions, and smart grid technology using automation, information and communication technologies (ICT), and artificial intelligence (AI). The primary objective is to optimize the generation, distribution, and consumption of energy, resulting in sustainable, cost-effective, and energy-efficient networks.

IES depend on a continuous, bidirectional exchange of information and electricity between utility providers and consumers. By leveraging advanced ICT, algorithms, and software, these systems interconnect multiple sectors—including electricity, gas, and heating grids—to effectively balance supply and demand. IES utilize adaptive controls and predictive AI to reduce unnecessary energy consumption in all settings, manage leading-edge technologies, and ensure safety and cybersecurity.  

Furthermore, IES empower users and utilities to shift energy usage away from peak pricing hours, resulting in substantial financial savings. They also facilitate localized energy generation and storage, enabling communities to maintain power even during broader grid outages. Finally, IES enhance sustainability by integrating variable energy sources, reducing environmental costs, and improving both economic and societal impacts.

Researchers

Chen, Nan
Fattahi, Javad
Habash, Riadh
Sun, Bo

Artificial intelligence is reshaping how we work, communicate, make decisions, understand, and move through the world. As these systems grow more capable and omnipresent, the ethical dimensions of their design, deployment, and governance become impossible to ignore. At uOttawa’s Faculty of Engineering, AI ethics, which is concerned with values like safety, fairness, transparency, accountability, and privacy, is not a peripheral concern for AI research. It is a foundational one.

In fundamental AI research, ethical considerations shape which problems are worth solving, how data is collected and used, and what risks are acceptable in pursuit of new knowledge. But the stakes are highest in applied AI research, where systems move from the lab into the real world, affecting real people in consequential domains: transportation, national security, healthcare, criminal justice, education, labour, and public infrastructure. Here, abstract values drive concrete design choices with measurable human impact.

Applied AI ethics demands that researchers ask not only can we build this? but should we, how, and for whom? It requires ongoing engagement with affected communities, rigorous testing for bias and other sociotechnical failure modes, and an honest reckoning with trade-offs between performance and competing values such as fairness and safety.

This work is increasingly inseparable from emerging law and policy. Landmark frameworks — including the EU AI Act and evolving standards from bodies like NIST and ISO — are translating ethical considerations into legal obligations. Applied AI researchers must understand and interact with these regulatory landscapes to build systems that are not only innovative but also lawful, trustworthy, and socially legitimate.

AI ethics is not a constraint on innovation; it is what makes innovation good.

Researchers

Bissessar, David
Fraser, Kathleen
Habash, Riadh
Hamou-Lhadj, Abdel
Intahchomphoo, Channarong
Millar, Jason

The Faculty of Engineering at the University of Ottawa is emerging as a national leader in applied AI research for autonomous systems, robotics, intelligent connectivity, and resilient cyber-physical infrastructures. The Faculty is home to an interdisciplinary ecosystem of researchers who specialize at the intersection of AI, robotics, communications, networking, control systems, sensing and cyber-physical security to advance next-generation autonomous and connected technologies.

Faculty researchers are actively developing technologies for intelligent transportation systems, uncrewed systems, swarm coordination, aerial base stations, autonomous decision-making, and AI-enabled situational awareness for dynamic and mission-critical environments. Research is supported by advanced laboratories, autonomous system testbeds, UAV platforms, immersive simulation environments, robotics and control facilities, and next-generation communication infrastructure.

The Faculty also plays a leadership role in large-scale collaborative initiatives focused on autonomous systems and AI-enabled infrastructures. This includes the Canada-wide “Training and Research in Reliable Services for Autonomous Systems in the Air and on Land” program and the Ontario-wide “Secure, Intelligent and Trustworthy Ecosystems for Connected and Autonomous Vehicles”  initiative, which together bring expertise from more than ten Faculty of Engineering researchers in communications, robotics, networking, autonomous systems, control, cybersecurity, and physical AI.

Faculty researchers also contribute to AI-enabled standards and standardization activities related to autonomous systems, including safety and operational frameworks such as UL, SOTIF, ISO initiatives, and 3GPP working groups to address intelligent and autonomous networked systems. The Faculty is closely integrated with the regional innovation ecosystem through collaborations with industry, government, and outdoor R&D facilities, enabling real-world experimentation and validation of autonomous systems.

Researchers

Bissessar, David
Briand, Lionel
Gueaieb, Wail
Kantarci, Burak
Laganiere, Robert
Lang, Jochen
Millar, Jason
Spinello, Davide

Intelligent sensing and measurement systems are sensor-based infrastructures that not only acquire physical quantities but also interpret, learn from, and act upon those measurements in a context-aware manner. 

Assisted by Artificial Intelligence (AI), they either fuse traditional transducers with embedded microprocessors and communication interfaces or adopt soft/virtual sensors. The integration of AI allows such systems to perceive patterns, learn models, analyze complex signals, and predict future states from continuous measurement streams. They deliver not only accurate estimates but also quantified uncertainty and traceable calibration, thereby supporting scalable, trustworthy measurements in highly automated and AI-enabled environments. This closes the loop from measurement to decision. 

Their wide spectrum of applications spans from smart factories and critical infrastructure to environmental monitoring, smart cities, medical measurements, and many others. In industrial production, they enable real-time process analytics, inline quality assurance, predictive maintenance, fault detection, and asset condition monitoring. For civil infrastructure, dense networks of intelligent nodes enable structural health monitoring of bridges, buildings, and nuclear facilities, detecting, localizing, and predicting damage. In environmental and urban applications, they measure water quality, air pollution, and other environmental parameters in real time, feeding smart-city control systems and supporting regulatory compliance and sustainability goals. In medical contexts, they enable continuous, real-time monitoring of physiological signals in hospitals, homes, and everyday life. Wearable smart sensors embedded in wristbands, patches, textiles, or earbuds track vital signs such as heart rate, respiratory rate, blood oxygen saturation, temperature, movement, and sleep, enabling early detection of deterioration and remote management of chronic conditions.

Researchers

Bissessar, David
Dragomirescu, Elena
Laganiere, Robert
Lang, Jochen
Matwin, Stan
Tee, Yi Hui
Yao, Jianping

As AI becomes deeply embedded in critical infrastructure, telecommunications, healthcare, and national security, cybersecurity and privacy present increasingly complex and urgent research challenges. At the Faculty of Engineering, a strong cluster of expertise addresses these challenges across the full spectrum — from theoretical foundations to operational deployment.

At the heart of this work is the uOttawa–IBM Cyber Range, a dedicated facility for simulating real-world cyberattack scenarios, enabling researchers, students, and industry partners to train and test defences under realistic conditions. Research conducted at the Cyber Range applies AI, large language models, and autonomous agents to problems such as phishing detection, malware analysis, vulnerability detection, intrusion detection, and log anomaly detection — with the overarching goal of neutralizing cybercrime before it can cause harm.

Complementary research addresses the challenge of building robust AI-driven security tools when labelled data is scarce or imbalanced, using generative models, semi-supervised learning, and anomaly detection to make threat identification more accurate and scalable.

On the IoT and embedded systems front, active research focuses on malware analysis, code fingerprinting, vulnerability assessment, and threat intelligence generation, applying AI to secure the interconnected devices that underpin modern infrastructure.

A growing area of focus is the security of AI systems themselves. As machine learning models are deployed in high-stakes settings, research examines threats such as data poisoning, adversarial attacks, and model manipulation — ensuring AI remains trustworthy even under deliberate attack.

Finally, theoretical work on cryptographic foundations — including post-quantum cryptography, pseudo randomness, and privacy-enhancing technologies — provides the rigorous underpinnings that secure AI-enabled systems depend on.

Researchers

Adams, Carlisle
Bissessar, David
Briand, Lionel
Chen, Nan
Hamou-Lhadj, Abdel
Jourdan, Guy-Vincent
Kantarci, Burak
Lu, Yiwei
Shirani, Paria

Researchers at uOttawa's Applied AI Center bring deep software engineering (SE) expertise to one of the field's most pressing challenges: making AI-powered software systems that practitioners can actually trust. Our work is guided by a core principle: AI is most effective when integrated into SE tasks with the same rigor and discipline that SE brings to any other systems.

We apply a range of AI techniques, including large language models (LLMs), natural language processing (NLP), and traditional machine learning, to improve the efficiency and reliability of software engineering activities, from requirements and design through implementation and testing. At the same time, we use SE techniques such as formal specifications, software modeling techniques, and software testing to strengthen the trustworthiness of AI-enabled systems, including autonomous systems with AI components.

Our current research spans reliable AI-based code generation, AI-driven requirements analysis, AI-assisted software design, testing, and verification of autonomous systems, and safety assurance of AI-enabled software.

By combining AI capabilities with established SE principles, our goal is to produce software systems that are not only more capable but also more reliable, trustworthy, and safe. We welcome collaboration with industry partners and government agencies seeking to adopt AI responsibly in their software practices.

Researchers

Bissessar, David
Briand, Lionel
Chen, Boqi
Lethbridge, Timothy
Shirani, Paria

Our professors bring a depth of knowledge in statistical and deep learning approaches to natural language processing (NLP) and computational linguistics. Specific areas of research include large language model (LLM) fine-tuning and evaluation, text classification, information retrieval, bias detection and mitigation, explainability, lexical semantics, and retrieval-augmented generation. We have a strong interest in responsible AI, including pre- and post-deployment safety evaluation.  

We also bring years of experience working in high-stakes domains such as healthcare and law, including partnerships with The Ottawa Hospital, Bruyere Research Institute, the Faculty of Law, and numerous Ottawa-based SMEs. Recent projects have included: the detection of hate speech in text and multimodal contexts; automated screening for mental, cognitive, and neurological disorders using patient-generated language; the identification of unintended risks associated with fine-tuning and retrieval in LLMs; and explainable emotion classification.  

In addition to core NLP research, our faculty is actively exploring applications of NLP to solve problems in science and engineering, including requirements engineering, automated test generation, regulatory compliance, and more.

Researchers

Fraser, Kathleen
Mao, Yongyi
Matwin, Stan
Rahgozar, Arya

Wireless networks can be considered the information highways of modern societies, providing connectivity for users, robots, and AI agents. Every generation of wireless networks, from 2G to 3G to 4G and finally 5G, has increased capacity, serving more bits to more users.  

In next-generation networks, AI will play a key role in enhancing performance and enabling new use cases. Innovations at the nexus of AI and wireless are gaining momentum. Using state-of-the-art AI techniques, adapting them to control problems of wireless networks, and providing innovative solutions that make networks better and more autonomous are at the heart of these innovations.  

The research opportunities allow graduate students to conduct award-winning research and earn prizes, as well as gain valuable industry experience through joint industry projects. In addition, elite programs such as the Canada Research Chairs, ongoing collaborations with the Vector Institute, and involvement in the activities of other premier AI institutes across Canada are providing excellent opportunities for students who want to excel in AI and wireless. This new AI center will amplify the impact of research, increase collaboration, and create new learning opportunities for students.

Researchers

Chen, Nan
Kantarci, Burak
Yao, Jianping

At the University of Ottawa, research on critical infrastructure focuses on enhancing the durability, resilience, and long-term performance of engineering systems through an integrated approach combining materials science, structural engineering, and data-driven methods. The work addresses fundamental degradation mechanisms and lifecycle challenges affecting infrastructure assets across diverse material systems and environments, while advancing sustainable, low-carbon solutions.  

Artificial intelligence is increasingly leveraged as a practical tool to better interpret complex datasets, support predictive modeling, and improve decision-making in infrastructure asset management. For example, AI-based approaches have been developed to automate advanced characterization techniques, including visual inspection and microscopic analyses, enhancing the speed, consistency, and objectivity of condition assessments.  

More broadly, these efforts aim to bridge experimental observations, field performance, and digital technologies, enabling more reliable, efficient, and sustainable management of critical infrastructure systems.

Researchers

Bissessar, David
Bonakdari, Hossein
Chen, Nan
Dragomirescu, Elena
Fattahi, Javad
Kantarci, Burak
Mohammadian, Majid
Sanchez Moretti, Leandro Francisco
Shirani, Paria
Wang, Di

Our team has established itself as pioneer in novel AI approaches applied to materials and fabrication. We put special emphasis on high performance, low cost and strong scalability for industrial production beyond the laboratory. Among others, we target key sectors including energy, aerospace, automotive, defence and manufacturing, with special focus on sustainability, hydrogen production and hydrogen storage.  

A differentiating specificity of our approach is its holistic nature where AI simulations, Density Function Theory (DFT) validation, experimental synthesis, laboratory state-of-the-art material characterization, property testing and performance evaluation are integrated in a sophisticated closed-loop to rapidly optimize both model performance and experimental outcomes in few iterations. This can lead to an exceptional reduction in the research and development cycles of complex materials, components, equipment and systems.  

A key strength of our approach is its focus on overcoming traditional limitations of experimental trials, empirical rules and first principle deterministic methods in navigating complex multielement systems.  

At the basic level, we build on big data and supercomputing to apply forward AI design for predicting properties of known materials and systems. At the advanced level, we have evolved into a world leader in inverse AI design for the discovery of novel materials and their fabrication processes for challenging applications with target properties and performance. Here, we predict and experimentally achieve optimal chemical composition, crystal structure, microstructure and synthesis parameters. This is achieved using unsupervised and/or semi-supervised models to integrate sequences of deconstructive AI algorithms including deep learning, evolutionary genetic learning, metric learning and adaptive learning.

Researchers

Fauteux-Lefebvre, Clémence
Jodoin, Bertrand
Karthikeyan, Adya
Lessard, Benoit
Sanchez Moretti, Leandro Francisco

Our current research activities in AI for healthcare span multiple areas, including human movement analysis, biomedical evidence processing, agentic healthcare systems, affective computing, and remote physiological sensing.

In human movement analysis, AI has permeated many areas, improving existing algorithms and revealing insights from large datasets. We are using AI for markerless motion capture, where skeletal joint angles are estimated directly from video. We are also using AI to build statistical shape models to distinguish healthy from pathological joints, with the goal of improving patient outcomes after osteochondroplasty. Finally, we are using evolutionary algorithms to generate predictive simulations of human gait augmented with wearable assistive devices.

Another important area of research focuses on applying AI to improve how scientific and clinical literature is searched, screened, extracted, synthesized, and continuously updated. By transforming large volumes of unstructured research into structured, actionable evidence, these approaches can help clinicians, researchers, and health-system decision-makers access current knowledge more efficiently, supporting evidence-based healthcare and reducing research waste.

We are also exploring agentic AI systems that serve as structured, human-centered collaborators, helping patients, clinicians, and health-system stakeholders navigate complex decisions and workflows. This includes conversational and recommender systems that support triage, care navigation, and context-sensitive guidance while preserving human oversight. A central emphasis is that healthcare AI agents should be explainable, safe, and socially accountable.

A further stream of work examines how intelligent systems can interpret human emotional and physiological states from language, behavior, wearable signals, and other multimodal data. The applications in this area include suicidal ideation detection using LLMs and synthetic data, remote physiological sensing, multimodal affect recognition, and AI-assisted mental health assessment.

Researchers

Bissessar, David
Fraser, Kathleen
Mao, Yongyi
Rahgozar, Arya
Turcotte, Marcel
Uchida, Thomas

Biopharmaceuticals, ranging from insulin to next-generation cancer therapies, are produced using living cells and are transforming modern healthcare. However, their manufacturing remains complex, resource-intensive, and costly. At the University of Ottawa, we are advancing machine learning approaches to improve the efficiency, scalability, and reliability of biomanufacturing processes.

Data-Driven Simulation: Traditional bioprocess optimization relies on running experiments under a limited set of conditions to identify those that maximize yield and minimize production time. These approaches often overlook the full dynamics of biological systems. By applying modern machine learning techniques to model complex time series data, we can build process-specific simulators that predict the entire trajectory of a bioprocess. This enables more accurate identification of optimal operating conditions and production timelines, improving equipment utilization and overall efficiency.

Scale-Up Modelling: Translating processes from laboratory to industrial scale is challenging due to changes in fluid flow, heat transfer, and mass transport. Instead of relying solely on theoretical models, we use deep learning trained on historical data to predict how processes behave at larger scales. This data-driven approach reduces the need for costly trial-and-error experiments and accelerates progress toward large-scale production of critical therapies.

Digital Twins: Modern bioreactors generate large quantities of sensor data. Machine learning models can convert this data into actionable insights, estimating key variables such as nutrient levels, cell density, and product concentration. Combined with predictive models, these systems form digital twins, i.e., virtual representations of the bioreactor, that enable real-time forecasting and automated control. This leads to higher yields, improved consistency, and reduced downtime.

Researchers

Chahal, Shawn