At the University of Waterloo’s Critical Machine Learning (ML) Lab, Dr. Sirisha Rambatra and her research team are working to build safe and efficient artificial intelligence (AI) models to advance fields such as medicine and aviation and help address the challenge of climate change.
The team’s research is based on fundamental machine learning (ML) theory to not only test the reliability of models, but also to predict and engineer them deeper. In practice, this means users can make safer and fairer decisions with lower computing costs and less waste.
Rambhatla, assistant professor in the School of Management Science and Engineering and the Val O’Donovan Professor of Efficient, Safe, and Adaptive AI, describes his work as “focused on building adaptive and responsible models that work well in the real world, not just the lab.”
“Whether it’s predicting flight delays or helping evaluate organ transplants, AI can help humans make faster and better decisions in critical situations,” she says.
Research for the real world
The institute’s research projects and collaborations span multiple industries and partners.
In health care, for example, the team is collaborating with Dr. Mamatha Bhat, a professor at the University of Toronto and a clinician-scientist at the University Health Network, Canada’s largest network of research and teaching hospitals, to improve the way clinicians match donor livers with transplant patients. The lab is using AI to develop tools that can suggest what’s best for specific recipients, potentially improving patient outcomes.
In aviation, the institute works closely with Navblue, the Waterloo-based Airbus company, to reduce flight delays. The company’s predictive models reportedly support smarter crew scheduling and improve delay performance by up to 60% on the most uncertain travel days, ultimately impacting not only carbon emissions but also employee and customer experience.
The team is collaborating with Apple on AI for Intelligent Production Monitoring, looking at how AI models trained in one environment can be adapted to work reliably in another, from healthcare and manufacturing to autonomous driving.
“Imagine building a self-driving system in sunny California and then using it in snowy Ontario,” Lambatra said. “Camera data is completely different. This is what we call a distribution shift, and we need AI that can handle this.”
Supported by an NSERC Discovery Grant, the lab’s research also focuses on improving representation learning and training ML models to “see” the world in different ways. For example, large feet cannot be automatically associated with a specific body size or shape. This helps generative AI modeling in healthcare provide more reliable patient diagnoses.
AI and data research with Dr. Sirisha Rambatra
Effective efficiency
An important aspect of the institute’s research is its approach to climate-smart AI. Training large models, such as those used to train large-scale language models (LLMS) like ChatGPT, can require significant computational resources, increasing both your carbon footprint and costs.
“We’re looking at ways to train these models faster on small graphics processing units (GPUs),” Rambhatla says. “Currently, building and using AI requires large amounts of water for cooling, and the energy used for training contributes significantly to emissions. If we can do this faster, we can significantly reduce the energy used in the process.”
A recent algorithm developed in our lab has already been shown to reduce LLM training time by 43% while maintaining accuracy. Projects like this also reflect broader goals around AI equity and accessibility, allowing smaller organizations with fewer resources to meaningfully contribute to generative AI without relying solely on a few companies with larger resources.
Industry collaboration and student opportunities
The Critical ML Lab conducts relevant and useful research, from algorithm design to testing to implementation.
“We don’t just build models in isolation,” says Rambhatla. “We co-develop them with our partners: clinicians, engineers, and front-line workers, so we can make sure they actually work, can be used, and are reliable.”
Dr. Sirisha Rambatra and students from the Critical Machine Learning Lab.
This collaborative, down-to-earth approach extends to the lab’s internal culture. Graduate students, undergraduate research assistants, and cooperative researchers are all actively involved in shaping the research agenda and engaging with industry collaborators.
Zhang Liu, a recently graduated master’s student and recipient of the University of Waterloo’s 2025 Alumni Gold Medal, worked on Rambhatra’s team and called it “meaningful work.”
“Most self-driving systems trained on clear-sky data fail in Canadian winters,” Liu said. “My work focused on improving these systems for snowy weather. The research was especially important because I knew my contribution could help prevent accidents and save lives.”
The lab’s emphasis on mentorship and practical impact makes it an attractive destination for students who want to push the boundaries of AI theory as well as ethical applications and social responsibility.
Looking to the future, Rambhatla hopes to expand the lab’s work on democratizing AI, allowing universities and even clinics to fine-tune powerful models without a supercomputer. “There is a huge gap between those who can use AI and those who can afford to build AI,” she says. “If we can lower that barrier, we can innovate safely and efficiently for everyone.”
Dr. Rambhatla’s work was recognized by the College of Engineering with the 2025 Engineering Research Excellence Award.
If you would like to learn about graduate research or industry partnership opportunities in the Critical Machine Learning Lab, please contact Dr. Sirisha Rambhatla.
Featured image: Dr. Sirisha Lambatra (front center) with students from Waterloo’s Institute for Critical Machine Learning. Photo credit @ Sam Chen.
