About me

I'm a PhD candidate at the University of Toronto, advised by Dr. Chris McIntosh and Dr. Michael Brudno. My research focuses on understanding why models fail in deployment and building methods to fix it: shortcut learning, spurious correlations, and distribution shift. While my primary application domain is healthcare, the methods are domain-agnostic and apply wherever robust generalization matters.

Before the PhD, I worked as an AI Research Engineer at A*STAR, Singapore, where I developed semi-supervised learning algorithms, collaborated with clinical partners, and generated IP that was commercially licensed. More recently, I've been working on LLM safety evaluation, RAG pipelines for large-scale document analysis, and multi-agent systems.

My research is supported by the Vector Institute, Schwartz Reisman Institute, University Health Network (UHN) and SickKids Hospital.

PhD Updates

• Our paper "Data-First Mitigation of Spatial and Spectral Shortcuts Without Introducing New Confounders" has been accepted at WACV 2026. This work presents a novel approach to addressing shortcut learning in AI models through data-centric methods rather than architectural changes or loss functions.
• Presented our work "When AI Democratizes Exploitation: LLM-Assisted Strategic Manipulation of Fair Division Algorithms" at the ACA Workshop, NeurIPS 2025.
• Excited to share that my research has crossed 600 citations on Google Scholar. Thank you to the research community for engaging with and building upon our work.
• Successfully achieved PhD candidacy. My thesis will be centered around addressing shortcut learning & improving model robustness for AI in healthcare.
• Excited to announce that I've been awarded the prestigious 2024 Schwartz Reisman Institute Graduate Fellowship. The fellowship supports my work on improving robustness and addressing shortcut learning in healthcare AI models.
• Our paper "Shortcut Learning in Medical AI Hinders Generalization" has been accepted at Nature Digital Medicine. We have also publicly released our code which allows researchers to estimate the generalization of their AI models without external data.
• I'm appearing as a guest on the ATGO-AI Podcast (Accountability, Trust, Governance, and Oversight of Artificial Intelligence) talking about my research and recent publication on shortcut learning and spurious correlations in healthcare data. The podcast is available on Spotify and Apple Podcasts: Episode 1 and Episode 2.

Publications

Volunteer Positions & Leadership

• Distinguished Reviewer: IEEE Transactions in Medical Imaging (TMI)
• Reviewer: Nature Digital Medicine
• Reviewer: Nature Scientific Reports
• Mentor: Graduate Application Assistance Program (GAAP)
• Chairperson: Singapore Computer Society (SCS) Chapter - NUS ISS
• Richard E Merwin Scholar - IEEE Computer Society
• Youth Excellence Award for Most Promising Engineer - Kerala
• Section Student Representative - IEEE Kerala Section
• Chairperson: IEEE Computer Society CET Chapter
• Co-Organizer: NASA Space Apps Challenge

Beyond Research

I love the sea, and I'm a certified scuba diver. P.S: If you ever need company to dive, drop me a message. Thanks to long layovers and commutes, I've been managing to read a bit. Here's my reading list if you are thinking "hmm, what should I read next?"