ML Engineer & Data Scientist

José P.
Barrantes

I build production ML systems, data platforms, and fraud prevention solutions that connect modeling to business decisions.

In practice, that means designing ML platforms, data pipelines, and decision-support systems that can be trusted in production. I choose the modeling approach that best fits the problem; whether that involves gradient boosting, Bayesian analysis, causal inference, deep learning, or a clear visualization.

Costa Rica
M.Sc. Computer Science · B.Sc. Biology
EN · ES
View Projects Resume PDF
José P. Barrantes
8+
Years in Data
50+
Projects Shipped
5
Industries
900+
Students Trained
Portfolio

Selected Projects

A collection of work across ML engineering, data science, and analytics.

Filter
MLOps Data Eng ML Eng

Feature Store for Time-Series Forecasting

2025 · Workshop @ DataDay Monterrey

Companion codebase for the DataDay 2025 workshop in Monterrey, Mexico: a working feature store for time-series forecasting, with Mage handling orchestration and Feast serving point-in-time correct features for training and inference.

Python Mage Feast Time Series Feature Store
Talk Page GitHub
Data Eng Data Science

High-Performance Analytics Pipeline

2025

End-to-end data pipeline on Iowa Liquor Sales (30M+ rows) covering the full data lifecycle: concurrent ingestion from the SODA API, transformation with Polars, persistent storage in DuckDB, geospatial analytics in R, and time-series forecasting with sktime. Accompanied by a hands-on tutorial article.

Data Engineering Mage Polars DuckDB R sktime
Read Article GitHub
ML Eng MLOps API

ML Model API Microservice

2024

Production-ready microservice predicting bike rentals with a CatBoost regressor. Features Pydantic request validation, automated unit/integration/inference tests with pytest, Docker containerization, and a CI/CD pipeline via GitHub Actions. Accompanied by a tutorial article on software engineering practices for data scientists.

FastAPI CatBoost Pydantic Docker pytest GitHub Actions
Read Article GitHub
Data Science Stats

Iowa Liquor Sales Analysis

2024

Deep-dive into 12 years of Iowa liquor sales (~28M invoices). Covers STL decomposition, isolation forest anomaly detection (via PyCaret), and Bayesian estimation (BEST) to answer two questions: does inventory diversity drive sales, and did the pandemic increase alcohol consumption? Built in R with Python interop through reticulate, published as a reproducible Quarto document.

R Time Series Bayesian Inference Anomaly Detection Quarto
Read Article GitHub
ML Eng Data Eng MLOps

Real-Time Fraud Prevention System

2025

End-to-end fraud prevention platform built as five Docker-composed microservices: a FastAPI data generator powered by the Synthetic Data Vault (SDV), Apache Kafka for event streaming, a Mage pipeline that consumes transactions and scores them with an ML model, and a Streamlit risk viewer (backed by DuckDB over Parquet) where analysts review high-risk cases.

Kafka Mage FastAPI Docker Compose Streamlit Fraud Detection
GitHub
Speaking

Talks & Workshops

Sharing knowledge at conferences across Latin America.

Conference Talk

Fraud Prevention, ML & Human-in-the-Loop Design

2026 · Nerdearla · Santiago, Chile 🇨🇱

Talk exploring the intersection of fraud prevention, machine learning, and human-in-the-loop software design patterns. The core idea: systems that keep analysts in the loop produce better outcomes than full automation.

Fraud Prevention ML ML-enabled Systems Software Architecture Design Patterns
Talk Materials Talk Page Nerdflix
Workshop

Feature Store for Time-Series Forecasting

2025 · DataDay · Monterrey, Mexico 🇲🇽

Hands-on workshop building a feature store for multi-series forecasting on Iowa Liquor Sales data. Covers ETL pipelines in Mage, feature definitions and point-in-time correct training datasets in Feast, DuckDB as the offline store, and online materialization to SQLite for model serving.

MLOps Feature Store Feast Mage DuckDB Time-Series Forecasting
Workshop Materials Talk Page
Toolkit

Skills & Technologies

Statistics & Analytics
Inference
Bayesian Data Analysis
Inference
Frequentist Inference
Inference
Causal Inference
Impact
Counterfactual Analysis
Time
Time Series Analysis & Forecasting
Design
Experimental Design & A/B Testing
Viz
Data Visualization
Viz
Data Storytelling
Practice
Uncertainty Quantification
Tools
Apache Superset
Machine Learning
Framework
Scikit-Learn
Framework
PyTorch
Framework
CatBoost
Method
Supervised Learning
Method
Gradient Boosting
Method
Time Series Forecasting
Method
Feature Engineering
Practice
Model Validation
Practice
Explainable / Interpretable ML
Practice
Model Evaluation & Error Analysis
Practice
Human-in-the-Loop ML
MLOps & Software Architecture
Architecture
ML Systems Architecture
Platform
ML Platform Design
Orchestration
Mage
Registry
MLflow
Architecture
Feast / Feature Stores
API
FastAPI
Containers
Docker
Infra
Kubernetes
CI/CD
GitHub Actions
Pattern
Human-in-the-Loop Systems
Linux & Infrastructure
OS
Linux
Shell
Bash / Zsh
VCS
Git
Container
Docker
Orchestration
Kubernetes
Cloud
IBM Cloud
Cloud
AWS
Config
Secrets & Environment Management
Fraud Prevention & Risk Systems
Domain
Fraud Detection & Analytics
Domain
Risk Scoring & Assessment
Pattern
Real-Time Detection Systems
Pattern
Human-in-the-Loop Review
Pattern
Decision Support Systems
Ops
Case Prioritization & Triage
Data Platforms & Data Engineering
SQL
PostgreSQL
OLAP
DuckDB
OLAP
ClickHouse
OLAP
Amazon Athena
Processing
Polars
Processing
Pandas / NumPy
Pipeline
ETL / ELT Pipelines
Pipeline
Feature Engineering Pipelines
Modeling
Data Modeling
Core Languages
Primary
Python
Primary
R / Tidyverse
Primary
SQL (PostgreSQL, ANSI)
Applied Domains
Finance
Fraud Prevention
Finance
Financial Risk
Industry
Fintech
Industry
Payments
Forecasting
Retail & Sales Analytics
Business
Marketing Analytics
Science
Computational Biology
Health
Healthcare & Clinical Data
Get in touch

Let's Connect

Open to collaborations, interesting problems, and good coffee.

Email Me GitHub LinkedIn