Human signals
Research
Machine learning in the space between signals and systems.
Language, speech, and clinical data become useful only when models are evaluated in the settings where people actually use them.
Themes
Four overlapping research directions
Clinical systems
AI in health
Machine learning in deployment, workflow, evidence, safety, and outcomes.Governance + evaluation
Responsible AI
Model behaviour, interpretability, fairness, privacy, standards, and public accountability.Place + network
Atlantic capacity
Building useful AI capacity in Atlanctic Canada, for students, collaborators, public institutions, and industry.Questions
Research questions worth making visible
Selected support
Recent grants and active work
Representation Immunization on Trainium: Scalable Noising & Weight-Locking
Responsible AI Amazon Research Award (ARA) · $257,710 (USD)
Preventing Cognitive Atrophy in LLM Mental Health Support Through Psychometric Evaluation and Therapist-Aligned Design
OpenAI, AI and Mental Health Grant Program · Co-investigator; Abeer Badawi: PI
A Unified Framework for the Safety, Interpretability, and Ethical Deployment of LLMs
NSERC Discovery · $345,000
Early identification of risk for major depressive disorder from parent and youth speech
CIHR Project · $1,009,800 · Co-applicant; Rudolf Uher: NPI
Research Grant
Center for AI Safety · $5750 (USD)
Partnership for Artificial Intelligence and Quality of Working Lives and Worker Well-Being (PAIQ)
SSHRC Partnership Grant - Stage 2 · $2,499,192 · Co-applicant; Arif Jetha: NPI