About
I’m an AI researcher focused on a question I keep coming back to: how fair, explainable, and aligned with human values are AI systems in real-world decisions — and how do we make them more so?
I work at the intersection of multi-agent systems, algorithmic fairness, and LLM-based value alignment, with a focus on making AI-assisted decisions more transparent and accountable — from fair, explainable AI decision-making for people affected by automated systems (published at ACM FAccT 2026) to adaptive AI models that learn and reflect people’s real priorities back to them.
What drives me most is building things that scale beyond a paper — work that actually reaches people and changes how they live or decide, not just what gets published.
I hold an MASc in Electrical and Computer Engineering (Pattern Analysis and Machine Intelligence) from the University of Waterloo, and I’m currently a Research Assistant at Waterloo’s Augmented Intelligence Lab.
Research interests
Multi-agent LLM systems, algorithmic recourse and counterfactual explanations, explainable and trustworthy AI, human value alignment, and evaluation of LLM-based systems.
Currently
Building an uncertainty-aware Bayesian model for value alignment, working with Prof. Edith Law at the Augmented Intelligence Lab, University of Waterloo. The system combines conversational, behavioral, and reflective evidence into an evolving model of a person’s priorities, with counterfactual previews that let people see how an action would shift the model’s interpretation before committing to it.
Previously
My MASc research, in the CHARM Lab under Prof. Amir-Hossein Karimi, focused on algorithmic recourse — when an AI system denies someone a loan or a job, what should they be told to do differently, and what happens when many people are competing for the same limited outcomes. That work is published at ACM FAccT 2026 and was also presented at the NeurIPS Workshop on Algorithmic Collective Action and at Remarkable 2026 (Vector Institute). I also built RecourseBench, a reproducible benchmarking platform spanning 22 methods and 137 configurations.
Before Waterloo, I worked on protein toxicity prediction (ToxTrans, awarded Best B.Sc. Thesis at Shahid Beheshti University) and built production search and ETL systems as a software engineer.
I’m currently looking for AI/ML research and engineering roles in Canada, open to Toronto. Feel free to reach out — my contact info is in the sidebar.
