Open to opportunities

Roberto Estrada

@robertoestrada1

Message

Data-focused analytics and credit-risk modeling specialist with hands-on SQL, Python, and explainable ML.

Mexico

Message

What I'm looking for

I’m looking for a role where I can build reliable, explainable ML and analytics with SQL/Python, validate data quality, and present insights to mixed audiences—ideally contributing to models that are production-ready and measurable (not just AUC-driven).

I’m a data-focused professional who builds practical analytics and ML systems with SQL and Python, with a strong emphasis on explainability. I’ve delivered end-to-end, SHAP-explainable credit risk modeling and I’m comfortable turning technical results into clear guidance for real decision-making.

In my credit risk project, I rebuilt multi-table aggregations in SQL/DuckDB and engineered domain features for the Home Credit Default Risk dataset. I also improved quality and reliability by monitoring data integrity, fixing calibration issues (including a predict_proba distortion from class_weight=balanced), and evaluating performance beyond AUC/KS using reliability curves, Brier score, and ECE.

I use model explanation tools to communicate both global and per-case drivers, and I frame thresholds with a cost-based FN/FP matrix rather than relying only on ranking metrics. I’ve served models via FastAPI, containerized with Docker, and added CI coverage with an 18-test GitHub Actions suite to keep the pipeline dependable.

Alongside building models and prototypes, I’ve taught and mentored programming and applied mathematics, designing hands-on projects that move learners from fundamentals to confident problem-solving. Through projects like early diabetes detection, employee turnover prediction, and flight delay prediction with PySpark, I aim for systems that are both accurate and easy to understand.

Experience

Work history, roles, and key accomplishments

Current

Mathematics & CS Instructor

Current

Algonova (by Algorithmics)

May 2026 - Present (3 months)

Designed and delivered programming and applied mathematics classes for young students, using high-performance international methodologies. Created instructional materials and hands-on projects to develop computational logic and progressive problem-solving.

Scratch Python Applied Mathematics Instructional Design Problem Solving

Instructor and Technical Advisor

Cunduacán, Tabasco

Aug 2022 - Dec 2024 (2 years 4 months)

Provided personalized instruction in algorithms and data structures using Python, applied to engineering problem-solving. Advised students on using NumPy, Pandas, and Matplotlib to visualize and analyze physical phenomena.

Python Algorithms Data Structures NumPy Pandas Matplotlib Scientific Visualization Data Analysis

Employee Turnover Predictor

Live App

Built and deployed a live Flask web application that predicts employee turnover using classification models. Designed it to help HR identify at-risk profiles for retention decisions.

Flask Machine Learning Classification

Credit Risk Scoring System

GitHub

Built an end-to-end probability-of-default credit risk model on the Home Credit Default Risk dataset, including multi-table SQL/DuckDB aggregations and feature engineering. Selected LightGBM vs Logistic Regression via cross-validation and served the model with FastAPI in a Docker container.

Python SQL DuckDB Pandas NumPy Feature Engineering Logistic Regression LightGBM scikit learn Brier Score Expected Calibration Error (ECE)SHAP fastAPI Docker GitHub Actions CI CD Cross Validation

Early Diabetes Detection

GitHub

Trained and compared multiple classification algorithms (including neural networks) for early detection of clinical pathologies. Performed a full Python/Pandas data cycle and produced Matplotlib/Seaborn visualizations to communicate findings and diagnostic features.