About me

I'm an associate lead data scientist at Octave in Colombo, where I primarily work on developing a machine learning based revenue management system for a hotel chain's dynamic pricing problem. Most of what I write here comes out of the work I do, my interests, and my previous work in policy research.

Before that#

I spent four years at Verité Research as a Lead Data Analyst where I conducted quantitative research on issues in the fields of socio-economics, politics, media, and macroeconomics. The research focus of this role has been invaluable in my work today. I am particularly proud of:

• Co-authoring the second country report the International Labour Organization has produced globally on the gender pay gap, using RIF regression to decompose where the gap actually comes from.
• Quantifying the value of unpaid care work in Sri Lanka at 42% of GDP — a number that has been spoken about in the Sri Lankan parliament when determining care policy.
• Developing a clustering algorithm that showed how Sri Lankan administrative boundaries could be realistically redrawn to hold a parliamentary majority with a minority of votes.
• Survey design and modelling on COVID vaccine uptake, air quality, and cigarette taxation.

Background#

MSc in Statistics from the London School of Economics, with coursework in deep learning, reinforcement learning, Bayesian methods and causal inference. Before that, a BSc in Mathematics and Economics from the University of London. I've also done specialist training in gender-sensitive macroeconomic modelling at the Levy Economics Institute and in care-extended CGE models at the National University of Mongolia and the University of Toronto.

More importantly#

I am extremely enthusiastic about complexity science and modeling complex economic systems. Many of my personal projects are geared towards building the tools to continuously create and improve complex models of the economy, integrating macroeconomic policy with its real effects on household wealth distributions and consumption inequality. I'm currently developing a five part toolchain under the name Syren for this very purpose with the end goal of having an optimised framework for developing economic ABMs, a state of the art calibration library, a graphics rendering library for visualising a simulation, a toolkit of agent based components that model detailed labour and household processes, and a software that glues these tools together.

What I write here#

Mostly developing ideas that still need work before I can turn them into more formalised pieces. If something here is wrong or you've solved the same problem differently, I'd genuinely like to hear about it.