Nairobi-born · statistics & modelling · Western University
I am Dhir — a quantitative researcher in training, vocalist, and model-builder. I care about what happens when a beautiful model meets a noisy world.
The question I keep returning to
How can we build models that remain useful when the world changes — and help people make better decisions?
Loading journey…

Plate 1 — Dhir Shah
My first real classroom was a seat on an Emirates 777 to Ahmedabad — a solo flight, a tiny notebook, a moving map. I tracked delay against connection buffer, headwinds against ETA drift, which immigration line moved faster and why. Without the words for it, I was already modeling a noisy system into states, constraints, and choices.
That instinct became a practice — explore, experiment, enact. It was forged the hard way: my first brush with proof-based calculus was a loud signal that curiosity alone wouldn't carry me, so I kept an error log, re-proved theorems from first principles, and stayed through office hours until the notation stopped feeling like rocket science.
Finance was a "no" for a long time — I come from a family of accountants, and the finance I knew was ledgers, not decision-making under uncertainty. Taking real hits on my own investments changed that: the path dependence I lived through wasn't in the textbook, but it was in the data. Prof. Bélisle's Financial Economics was the bridge, and it left me a test I still use — if a non-quant can't act on your analysis, you haven't finished the job.
// I am not applying to check a box. I am chasing the intersections that first came into focus at 35,000 feet.
LEG 01
University of Toronto
Double Major: Economics & Mathematics · Minor: Statistics, with Distinction
Sep 2021 — Jun 2026 · conferred
Arts & Science Internship (Co-op). TA in Mathematics (MAT235Y1) and Financial Economics (ECO358H1).
Manifest — coursework
LEG 02
Western University
Specialization: Financial Modelling
Sep 2026 — present
Western Graduate Research Scholarship recipient. Graduate Teaching Assistant for SS2244: Statistics for Science. Studying stochastic modelling and financial econometrics with a focus on models that hold up out of sample.
Manifest — in progress
“So the story changes when the order changes.”
F-01 / Quantitative FinanceA simulation-based approach to portfolio risk
Consolidated two years of market data from Google Finance to compute daily drift and volatility, then ran 10-day price evolutions to deliver a 99% VaR with 95% historical accuracy. Built a 10-day Monte Carlo VaR model for an $11.9B, 7-stock portfolio in Excel, using Cholesky decomposition in STATA to cut simulation time by 70%. Automated the multi-step simulation with correlated random draws, enabling 10,000 daily recalculations in under two minutes.
My work is moving toward a broader but sharper question: how do we build statistical and financial models that remain useful when conditions change — and help people make better decisions when the stakes are real? I care about the full chain: what the model assumes, what it finds, when it fails, and whether someone can understand and act on the result.
I want to understand what remains useful when economic conditions, behaviour, and the data-generating process shift. A good score on yesterday's data is only the beginning.
Neural networks and flexible optimizers can search spaces far larger than any person can hold in their head. I am interested in the mathematics underneath: what they find, and when flexibility becomes overfitting.
The final test is not elegance alone. Can a manager, student, or institution understand the assumptions, question the output, and make a better decision when the stakes are real?
Formal training gave me an ear for structure and improvisation — how a raga constrains choice but expands expression. After an OCD diagnosis, it's also how I regulate attention, and a standing reminder that every model I build ultimately serves the person using it.
My own composition in Raga Yaman.
A traditional Tirvat composition in Raga Bhupali.
A traditional bandish in Raga Durga, set to Jhaptaal.
My rendition of Bol Do Na Zara, originally by Armaan Malik and composed by Amaal Malik.
Watch on YouTubeA group performance dedicated to graduating friends, woven from songs about friendship.
Watch on YouTubeA performance at the academy's global meet, before Shankar Mahadevan himself (from 1:47:28).
Watch on FacebookOpen to research collaborations in financial econometrics and risk — a good problem, a supervisor lead, or a note on a raga. Pick any channel below and I'll fly it back.
Contact
Use the secure form