Churn Radar
A gradient-boosted model that flags at-risk customers 30 days out, with SHAP explanations baked in so the retention team knows exactly why — not just who.
Available for new projects
Data scientist turning messy data into decisions that hold up.
I'm Alex Rivera — I build models, pipelines, and dashboards that survive contact with real data. Currently working on demand forecasting at Northwind Labs.
import pandas as pd
df = pd.read_sql(query, conn)
# clean, then model
df = (df
.dropna(subset=["churned"])
.pipe(engineer_features)
)
model.fit(X_train, y_train)
print(f"AUC: {auc:.3f}")
01 — About
A bit about how I work
Five years turning raw data into models, dashboards, and decisions that actually get used.
I'm a data scientist who got into this field because I like the moment a messy dataset turns into a clear answer. Most of my time goes into building models, designing experiments, and the pipelines that keep both of those honest.
Before that I spent a couple of years doing data analysis and BI work, which left me with a healthy obsession over where numbers actually come from. I try to build models and dashboards that the next analyst can trust and extend — not a black box they're afraid to touch.
Outside of work I write about applied statistics, contribute to a couple of open-source data tools, and am slowly building a personal project to make A/B test analysis less error-prone for small teams.
Languages
Python
SQL
R
Scala
ML & Stats
scikit-learn
PyTorch
XGBoost
statsmodels
Data Infra
Spark
Airflow
dbt
Snowflake
Viz & Tooling
Tableau
Plotly
Jupyter
MLflow
02 — Projects
Things I've built
A mix of production models, open-source tools, and analyses that changed a real decision.
Significant
An open-source A/B test analysis tool that catches the usual traps — peeking, underpowered samples, multiple comparisons — before a team ships a false win.
Demand Forecaster
A production forecasting pipeline blending a seasonal model with gradient boosting on holidays and promotions, cutting weekly stockouts by 23%.
Metrics Layer
A dbt-based semantic layer that gives every team the same definition of "active user" and "revenue" — ending the recurring fights over whose dashboard is right.
03 — Timeline
Where I've been
Roles, in order, with the parts that mattered.
2024 — PresentCurrent
Senior Data Scientist
Northwind Labs
Own demand forecasting and the experimentation platform. Rebuilt the team's forecasting pipeline, cutting weekly stockouts by 23% and forecast error by a third.
2022 — 2024
Data Scientist
Fieldstone
Built the churn prediction model and retention scoring system from scratch, surfacing at-risk accounts 30 days earlier with clear, explainable drivers.
2020 — 2022
Data Analyst
Loop & Co.
Owned reporting and ad-hoc analysis for the growth team — built the first version of the company's metrics dashboard. Where I learned that a trusted number beats a clever model nobody checks.
2019
B.Sc. Statistics
University of Lyrth
Focused on applied statistics and machine learning. Senior project: a Bayesian model for predicting transit delays, later adopted by the city's planning office.
Have something worth building?
I'm open to data science and analytics roles and the occasional contract project. The fastest way to reach me is email.