I'm the first data scientist at a consulting firm, working across sales, marketing and operations. I turn messy Salesforce data and half-documented business processes into forecasts, models and reporting systems people actually use.
Currently
Data Scientist, ABGi USA
Focus
Forecasting · attribution · machine learning
01 — Marketing attribution
10.3×
blended ROAS, surfaced across three channels
Campaign tools tracked clicks and booth scans. Salesforce tracked signed contracts. Nothing connected them, so no one could say which activity produced revenue. I built the identity-resolution layer matching contacts to accounts, the touchpoint table underneath it, and the report leadership now uses at budget time.
02 — Revenue forecasting
100+
active clients billed from the engine each month
Revenue lands when projects clear billing triggers, and the four contract types clear on two entirely different logics. The forecast was being reconstructed by hand in spreadsheets, one client at a time. The execs and I modeled each contract type under its own rules, estimated milestone slippage from how projects historically completed rather than from CRM dates, and built the engine finance now bills from.
03 — Retail analytics
0.77
held-out R² predicting full-season sell-through
Merchandisers commit to reorders before they know how a style will sell. Week-one velocity turns out to be useless for predicting the season, because a large buy mechanically produces large velocity. Dividing units sold by units bought lifts the correlation from 0.07 to 0.75, and a linear model on that predicts full-season sell-through within about eight points.
My background is in quantitative economics (M.A., UCLA). Most of the work I do starts with a business question that has already been answered three different ways by three different people.
I'm the first data scientist at a 77-person consulting firm, so the job runs end to end. I pull data out of systems that weren't built to give it up, write down rules that were never written down, build the model, then build the thing people open on Monday morning.
Outside work I paint, follow fashion, and build small apps. I care how these pages look because reports that are pleasant to read get opened more often.
ABGi USA
Data Scientist
Apr 2025 — present · Houston
First data scientist hired by a 77-person consulting firm. Built the forecasting, attribution, and executive reporting systems now used across sales, marketing, finance, and operations.
Data Gals Initiative
Founder & Mentor
Sep 2024 — present
A mentor community for women entering data. Focused on getting people to finished projects, technical confidence, and the parts of career navigation nobody explains.
ABGi USA
Data Science Intern, Real Estate Practice
Jun — Dec 2024
Joined as the company's first data science hire of any kind and converted to full time. Early work on reporting and analysis for the real estate practice.
Optimas Capital
Quantitative Analyst
Jan — Mar 2024
Research on systematic strategies: reduced a 2,000-feature financial dataset with PCA, then modeled signals with random forest and gradient boosting.
Equity Methods LLC
Data Analyst
Aug — Sep 2023
Valuation and equity compensation analytics, working with the data behind client reporting.
UCLA
M.A. Quantitative Economics
Graduated Mar 2025 · GPA 3.86
Econometrics, causal inference, and machine learning. B.A. Economics, University of Delhi, before that.
Some of this was done independently, some with Optimas Capital. Notebooks exclude confidential data.
Analysed a marketing A/B test across 580,000 users: data cleaning, exploratory analysis, and Chi-Squared testing for significance, segmented by day and hour.
Reduced 2,000+ financial features with PCA, selected on Information Coefficient, then modelled CSI 800 returns with random forest and XGBoost to find alphas uncorrelated with the market.
Scored two decades of New York Times coverage for sentiment, converted the scores into currency signals with a dynamic Kalman filter, and backtested against the HFRI Currency Index.
Fine-tuned BERT on 2020 election tweets and tested how stance toward political figures changed the classification.