About Me

Most ML systems tell you what happened. Very few tell you why, and even fewer change what you actually do about it.

That gap between correlation and causation costs organizations millions in misallocated decisions every year, whether in financial institutions, marketing teams, or recommendation systems. I have spent the last 3 years focused on one question: how do you move from observational data to decisions you can actually trust?

At TD Bank, I was part of a 6-person team that built the end-to-end causal ML pipeline for mortgage renewal pricing, from CATE model development and customer segmentation through to LP optimization and production rate delivery, landing $28M in incremental value vs. baseline.

Before that, at Shomara, I used GRF-based CATE estimators and CUPED variance reduction to measure the causal effect of credit on purchase behavior, one of the earlier applications of heterogeneous treatment effects in fintech.

Alongside the industry work, I completed a Master’s in Applied Mathematics at UBC with a 4.0 GPA, designing an algorithm using Optimal Transport to jointly estimate drift, diffusion, and causal structure from temporal snapshots, submitted to JMLR and available on Arxiv. I also built a randomized experiment framework to measure how individual prompt words causally shift LLM outputs.

My research interests lie at the intersection of Causal Machine Learning and Generative AI, focusing on building interpretable AI systems that understand cause-and-effect relationships rather than mere correlations.

I like working at the edge where rigorous math meets real decisions. If you are working on hard causal problems in industry or research, I would love to connect.

Education

Master’s in Applied Mathematics, University of British Columbia (2023 - 2025) · GPA: 4.0/4.0 Thesis: Identifying drift, diffusion, and causal structure from temporal snapshots First comprehensive approach for jointly estimating drift, diffusion, and causal structure of SDEs from temporal marginals using entropic optimal transport.

Bachelor’s in Computer Science, Sharif University of Technology (2019 - 2023) · GPA: 4.0/4.0 Foundation in algorithms, probability, statistical learning, and software engineering with perfect grades in core courses.

Research & Industry Experience

Applied ML Scientist 1 at TD Bank (Nov 2025 - Present)

◦ Built the end-to-end causal ML pricing pipeline for mortgage renewals, from CATE model development and customer segmentation to rate optimization, delivering $28M in incremental value vs. baseline with a 6-person team

◦ Worked closely with business stakeholders to translate model outputs into real pricing rates, aligning on requirements and pushing toward production

◦ Led knowledge transfer on project completion, documenting the full pipeline and onboarding the team before transitioning to a new mandate

◦ Currently contributing to foundational causal inference research with Layer 6, building a proof-of-concept model that approaches baseline performance on AUC for GIC pricing task.

Research on LLM Attribution with Prof. Mathias Lecuyer, UBC (Sep 2024 - Present)

Designed a randomized experiment framework to measure the causal attribution of prompt words on Large Language Model outputs, estimating Average Marginal Effects (AME) of individual tokens.

Master’s Research with Prof. Elina Robeva, UBC (Sep 2023 - Present)

Developed APPEX, an iterative algorithm using Optimal Transport to jointly estimate drift, diffusion, and causal graphs from temporal snapshots of Stochastic Differential Equations. Preprint on arXiv, submitted to JMLR.

Data Scientist at Shomara (Dec 2021 - Oct 2022)

Applied causal inference in fintech: Estimated Conditional Average Treatment Effects using GRF with novel experimental designs and CUPED variance reduction. Improved default prediction models using XGBoost, achieving 5% relative AUC increase.

Data Science Intern at Snapp Market (Sep 2021 - Dec 2021)

Developed computer vision algorithm improving staff location tracking by 20% using YOLO and OpenCV. Built SQL-based marketing dashboards for business intelligence.