Data Scientist specializing in Causal Inference

Engineer in statistics and data analysis with deep expertise in causal inference, econometrics, and machine learning. I build rigorous statistical methods to answer hard business questions: not just what happened, but why. Specialized in instrumental variables, synthetic controls, and difference-in-differences estimation. Passionate about turning complex data into actionable insights for product, analytics, and research teams.

View Projects GitHub

Featured Projects

Causal Uplift with Double ML IV

Estimated the causal effect of ad exposure on user visits using Double ML with an Instrumental Variable (DMLIV) on the Criteo incrementality dataset (25M rows).

Key Finding: Decomposed OLS bias into 66% confounding + 34% endogeneity, revealing 210.8% overstatement vs causal estimate
Double ML Instrumental Variables Python Econometrics Large-scale Data

Regression Discontinuity Design: FICO 675 Threshold

Estimated the causal effect of crossing the FICO 675 credit score threshold on approved loan amounts using sharp regression discontinuity on Lending Club data (2.26M loans). Rigorous assumption testing with McCrary density, covariate smoothness, bandwidth sensitivity, and placebo cutoffs.

Key Finding: Crossing FICO 675 causes $256 increase in approved loan amount (95% CI: [$220, $293]). Placebo test at 650 revealed $3,410 effect, exposing multiple approval rules and limiting generalizability.
Regression Discontinuity Sharp RDD Python Assumption Testing Policy Evaluation

Staggered Difference-in-Differences (Callaway & Sant'Anna)

Estimated heterogeneous treatment effects in staggered adoption settings using the Callaway & Sant'Anna (2020) estimator, addressing TWFE bias with treatment effect heterogeneity.

Key Finding: TWFE and CS estimators converged (~7.88pp vs ~8.02pp) because effects were homogeneous—identified that TWFE bias manifests with heterogeneous effects only
Staggered DiD Callaway & Sant'Anna Python Panel Data Causal Effects

Instrumental Variables & 2SLS: Causal Returns to Education

Estimated the true causal effect of education on earnings using Two-Stage Least Squares with college proximity as an exogenous instrument on Card (1995) data (3,010 men).

Key Finding: IV revealed 14.5% return vs 7.5% OLS estimate—diagnosing 93.9% ability bias in naive regression, showing how uncontrolled confounding inflates estimates by nearly 2x
2SLS Instrumental Variables Econometrics Python Exogeneity Testing

Synthetic Control Methods

Estimated causal treatment effects in observational settings using synthetic control methodology, creating weighted combinations of control units to match the treated unit's pre-treatment trajectory.

Key Finding: Synthetic control revealed treatment effect magnitude by comparing observed post-treatment outcome to the counterfactual trajectory of the synthetic control unit
Synthetic Control Time Series Python Causal Inference Policy Evaluation

Technical Stack

Causal Inference

  • Instrumental Variables
  • Double Machine Learning
  • Difference-in-Differences
  • Synthetic Control
  • Regression Discontinuity

Econometrics

  • 2SLS / IV Estimation
  • Panel Data Methods
  • Endogeneity Testing
  • Assumption Validation
  • Treatment Effects

Programming

  • Python (pandas, numpy)
  • Statsmodels
  • Matplotlib / Plotting
  • Git / GitHub
  • Data Processing

Domains

  • Labor Economics
  • Product Analytics
  • Policy Evaluation
  • A/B Testing
  • Large-scale Data

Resume

Download my full resume to see my professional background and experience

Download Resume

Get in Touch