01 — Descriptive
Make patterns clear
Compelling visualizations, interaction matrices, and storytelling frameworks that make patterns instantly understandable.
BI Engineer · Causal & Decision Intelligence
Turning complex data into confident decisions
I design and build robust data systems that turn raw data into reliable foundations for advanced causal analysis — and transform complex datasets into clear, actionable strategic insights.
About
I'm a BI Engineer with deep expertise in Causal & Decision Intelligence. My work spans three interconnected layers of analytics — from making patterns understandable, to explaining why outcomes occur, to recommending concrete actions.
02 — Diagnostic
Explain why
Rigorous causal inference — DAG construction, Bayesian statistics, and counterfactual analysis — to uncover true root causes.
03 — Prescriptive
Recommend action
Game theory, optimization, and strategic modeling to recommend concrete, measurable actions.
Currently
BI Engineer at Meet5
I build end-to-end analytical infrastructure — BigQuery pipelines, retention & revenue tracking systems, and interactive Dash applications — that powers both operational reporting and high-impact causal research. I also work in geospatial analysis (H3, Kepler.gl) and complex network analysis (NetworkX). Beyond technical execution, I'm committed to mentoring and elevating analytical thinking, combining scientific rigor — inspired by Popper's falsifiability and Pearl's causal framework — with practical business acumen to move teams from data confusion to confident, evidence-based decisions.
Featured Projects
Working paper · Live app
Compliance-Gated Data & Selection Bias
It starts simply — predict the greenhouse-gas emissions of Chicago's buildings — but turns into a case study in how understanding the Data Generation Process (DGP) solves structural missing data caused by selection bias. Only buildings larger than 50,000 sqft are required to report, so the observed data isn't a representative sample but a biased sample of those that complied. The bias lives in the DGP itself, so no better model or pipeline corrects it unless it integrates the DGP.
The broader idea: much of our data isn't a neutral snapshot — it's constituted by a documented process (a regulation, an accounting standard, an eligibility rule). That document doesn't just describe the data, it creates it — so knowledge of the DGP can be extracted from these documents and built into the analytical pipeline. I demonstrate this with structural causal models, a working paper, and an interactive app.
A developing idea, shared to think it through in the open — feedback and critique on the framing or method are genuinely welcome.
Excel
Python
PostgreSQL
