$ data scientist & BI analyst
turning messy, multi-source data into decisions you can act on!
I am a data scientist with 5+ years of experience with dual Master's degrees, working at the intersection of
experimentation and causal inference. I design and read A/B tests, build predictive models in R and Python,
and create Power BI dashboards that let stakeholders answer their own questions.
Selected academic and industry work — from causal survey experiments to production analytics and executive dashboards. Each card opens a detailed write-up.
>industry projects
Applied analytics and business-intelligence work — behavioral operations, market segmentation, dashboards, forecasting, and survival modeling.
causal inference · advertising incrementality
Did the Ads Actually Work? Incrementality from a 588k-User Experiment
Ads vs. PSA control, read three ways: an exact permutation test (1M label
shuffles), Bayesian posteriors that turn lift into decision probabilities, and a
from-scratch T-learner whose Qini curve turns the average effect into a media
plan — with two interactive widgets pricing the campaign under uncertainty.
PythonNumPyPermutation TestBayesian A/BUplift · QiniInteractive ×2
view case study + dashboard →
behavioral operations · coordination economics
T-Baar: Reducing Empty Returns in Iran's Road-Freight Network
Diagnosed empty returns (~30% of capacity) as a coordination failure, then built a
bipartite matching model — quantified on 11,594 real trips (−53% empty km) — with an
interactive matching-adoption simulator.
PythonOptimizationBehavioral EconInteractive
view case study + dashboard →
unsupervised ML · market segmentation
Canadian Freight Corridor Segmentation: Five Market Archetypes from CFAF Data
Full unsupervised pipeline on 58,721 Statistics Canada CFAF records — 13-feature
engineering, PCA (5 PCs → 90% variance), K-Means (k=5) validated by silhouette
and Ward dendrogram, t-SNE map — segmenting 194 corridors into five interpretable
archetypes that feed T-Baar's demand calibration.
Pythonscikit-learnPCA · K-MeansInteractive
view case study + explorer →
business intelligence · dashboards
Executive KPI Dashboards for Demand & Supply Risk
Self-serve Power BI dashboards over a star-schema model, with an OLS trend forecast
(95% band) that informed a capacity decision — plus an interactive KPI explorer that
recomputes by metric and load type.
Power BISQLForecastingInteractive
view case study + dashboard →
survival analysis · event history
Time to Event: Survival Modeling of Lung-Cancer Outcomes
Kaplan-Meier curves, a Weibull accelerated-failure model, and a Cox
proportional-hazards model on the NCCTG lung cohort — all built from the
likelihood up, including the Cox partial likelihood with Efron ties.
Reproduces survreg and coxph exactly.
RPythonCox PHCensored MLE
view case study →
>academic projects
Experimental and causal research in political science and international relations.
survey experiment · causal inference
Framing War: Partisan Identity & Public Support for Military Interventions
A pre-registered 5-condition survey experiment (N = 2,350) testing how elite cues
and humanitarian-vs-security framing causally shape support for U.S. interventions.
Ordinal & multinomial logistic regression; includes an interactive deck.
RRCT DesignOrdinal/Multinomial LogitQualtrics
view case study + deck →
multinomial logit · MLE
An Executive Glass Ceiling: Gender & the Careers of Mexican Legislators
A from-scratch multinomial logit on ~1,900 legislators' first career moves. Likelihood,
gradient and marginal effects.
RPythonMLEMarginal Effects
view case study →
survival analysis · competing risks
How Long Do Governments Survive? Coalition Complexity & Cabinet Collapse
Kaplan-Meier, a from-scratch Cox model (Efron ties), a Weibull AFT, and a
competing-risks decomposition on ~1,500 ParlGov cabinets — plus a
dependency-free interactive dashboard that predicts a cabinet's lifespan live.
PythonNumPyCox PHWeibull AFTInteractive
view case study + dashboard →
count models · poisson & negative binomial
Petrostates & Conflict Initiation: Coup Risk, Oil & Diversionary War
My own research, rebuilt from scratch: Poisson (IRLS) and negative-binomial
MLE on ~5,600 country-years, modeling how coup risk and oil wealth drive
militarized disputes — with an interactive dashboard for the key interaction.
PythonNumPyPoissonNegative BinomialInteractive
view case study + dashboard →
Founder & Analytics Lead
T-Baar Freight Analytics
May 2025 – Present
Tehran, Iran
- Conducting market research on Iran's road-freight sector to reduce empty returns, unlocking ≈$300,000 in annual savings per corridor.
- Designing bipartite load–driver matching algorithms and predictive models to increase utilization and improve on-time performance.
- Planning a 90-day pilot with digital waybills, escrow-based payments, and e-queuing to demonstrate multi-million-dollar economic and environmental benefits.
Consultant — Data & Business Intelligence (Part-time)
Afagh Zarrin BD Co.
Jun 2025 – Dec 2025
Tehran, Iran
- Re-architected reporting and forecasting pipelines; built demand models that informed a major capacity investment and capital allocation decision.
- Deployed executive dashboards for volume, margin, and supply-risk KPIs, replacing spreadsheet reporting with near real-time monitoring.
Teaching Assistant (Full-time)
Department of Political Science, Rutgers University
Sep 2024 – Apr 2025
New Brunswick, NJ, USA
- Assisted instruction in courses on statistical analysis, experimental design, and causal inference.
- Guided students through hands-on R labs for data cleaning, wrangling, modeling, and visualization.
- Graded assignments and provided targeted feedback on coding practices and quantitative communication.
Summer Fellow (Hybrid)
National Union for Democracy in Iran (NUFDI)
Jun 2024 – Sep 2024
Washington, DC, USA
- Designed and programmed survey experiments to measure public sentiment on policy issues.
- Analyzed experimental data in R to generate statistically robust insights for advocacy and communications.
Excellence Fellow (Full-time)
Department of Political Science, Rutgers University
Sep 2022 – Aug 2024
New Brunswick, NJ, USA
- Led projects using Bayesian methods to analyze voting behavior and preference formation.
- Applied advanced regression models to predict categorical outcomes from survey data.
- Designed and executed randomized controlled trials and A/B-type experiments from design through impact evaluation.
Research Assistant (Part-time)
Department of Political Science, Rutgers University
Jun 2023 – Sep 2023
New Brunswick, NJ, USA
- Employed maximum likelihood estimation (MLE) to estimate causal effects in experimental data, including diagnostics and robustness checks.
- Developed statistical models and publication-ready visualizations in R to support peer-reviewed research outputs.
- Performed multivariate regression to analyze how demographics, ideology, and media exposure affect behavioral outcomes.
Data Analyst (Part-time)
Afagh Zarrin BD Co.
Jul 2018 – Dec 2020
Tehran, Iran
- Built analytical tables from multi-year transaction, pricing, and logistics data, enabling consistent demand and profitability analysis by product and corridor.
- Produced recurring performance reports and visuals that shifted management discussions from intuition to data-backed pricing and capacity choices.
Junior Data Analyst (Part-time)
Afagh Zarrin BD Co.
Feb 2016 – Jul 2018
Tehran, Iran
- Cleaned and reconciled sales and inventory data, reducing reporting errors and providing the firm’s first consistent KPI views for leadership.
Commodity Manager Assistant (Part-time)
Tehran Urban & Suburban Railway Operation Co.
Jun 2013 – Jun 2014
Tehran, Iran
- Managed supplier relations and analyzed market data to identify investment and procurement opportunities.
Sales Manager Assistant (Part-time)
Zagrutti Men's Clothing
Sep 2011 – Mar 2012
Tehran, Iran
- Oversaw sales staff, resolved customer complaints, and supported pricing strategies to optimize revenue.
Production Assistant (Part-time)
PARS Ceram Company
Sep 2009 – Mar 2010
Qarchak, Tehran, Iran
- Maintained production databases and managed documentation workflows to support traceability and quality control.
M.A. Political Science & Government
Rutgers University
2022 – 2025
New Brunswick, NJ, USA
M.A. International Relations
University of Tehran
2015 – 2020
Tehran, Iran
B.S. Industrial Management
Allameh Tabataba'i University
2009 – 2014
Tehran, Iran
I work end-to-end: designing and reading A/B tests and randomized experiments from hypothesis and power analysis through causal effect estimation; building predictive models in R and Python on top of production-grade SQL; and turning results into self-serve Power BI dashboards. Across projects, the constant is translating rigorous analysis into clear, honest recommendations — for technical and non-technical audiences alike.
> data analysis
Predictive Modeling
Generalized Linear Models (GLM)
Maximum Likelihood Estimation (MLE)
Logistic/Probit Regression
Time-Series Analysis
Hypothesis Testing
> programming languages
R (advanced)
Python (intermediate)
SQL (intermediate)
> business intelligence
Power BI
Dashboard Creation
KPI Tracking
Automated Reporting
ArcGIS
> experimental design
A/B Testing
Randomized Controlled Trials (RCTs)
Causal Inference
Regression Discontinuity (RDD)
Survey Design (Qualtrics)