Data Scientist · DGII, Dominican Republic

Raudy García outlier · 3.2σ

I build models that find the taxpayer, transaction, or pattern that doesn't belong — anomaly detection and predictive modeling for the Dominican Republic's tax authority, and a growing body of independent ML work aimed at fintech.

Location — Bonao, Dominican Republic
Stack — Python · SQL · scikit-learn
Focus — ML engineering & fintech
Taxpayer risk distribution — illustrative n = 240
low — risk signal — high ● flagged outlier
About

From technical support desks to fiscal risk models.

I started in Bonao fixing hardware and walking people through unfamiliar software. Since then I've moved through banking analytics and into the Dominican Republic's tax authority, where I now build the models that decide which returns deserve a second look.

What hasn't changed is the instinct: take a pile of messy, high-volume data and find the thing in it that actually matters — the pattern, the leak, the outlier. I'm currently pushing that instinct further, working toward an ML engineering role in fintech, with GCP certifications and independent projects (sports prediction, anomaly detection, forecasting) alongside my day-to-day work.

Currently Data Science, Dirección General de Impuestos Internos (DGII) — the Dominican Republic's national tax authority
Core stack Python (pandas, scikit-learn, XGBoost, SHAP), SQL Server, Excel/openpyxl
Working toward ML engineering / fintech roles, GCP Cloud Digital Leader → ML Engineer certification track
Education B.S. Computer Systems Engineering, UCATECI (2019–2025)
Experience

Where the work happened.

03/2026 — Present

Data Science

Dirección General de Impuestos Internos (DGII) · Santo Domingo

The Dominican Republic's national tax authority, responsible for administering, regulating, and collecting all internal taxes nationwide.

  • Formulate and implement predictive models and anomaly-detection algorithms to identify tax evasion patterns across high-volume fiscal datasets.
  • Process, clean, and model complex data structures using SQL and Python, turning raw tax data into actionable insight.
  • Design analytical reports that strengthen institutional oversight and strategic decision-making.
08/2023 — 02/2026

Data Analyst

Banco Agrícola
  • Analyzed data to generate reports and surface trends, patterns, and insights across the bank's operations.
  • Built and maintained Python applications for data extraction and automation, improving workflow efficiency.
  • Deployed PostgreSQL via Docker for consistent development and testing environments.
  • Built web-scraping pipelines to collect structured data from multiple online sources.
  • Designed and deployed REST APIs (Django REST Framework, Flask) exposing structured datasets for internal analytics.
05/2021 — 08/2023

Technical Support

Ministry of Education · Bonao, Dominican Republic
  • Resolved hardware and software issues to keep systems stable and performant.
  • Provided hands-on support and program configuration for over 50 users.
Projects

Applied work, on the clock and off it.

Client and employer projects on the left; independent, open-source work — including the GCP-track and fintech-facing builds — below.

Internal Dashboard for Loan Monitoring

Banco Agrícola
TechPower BI, SQL Server, Excel
RoleBuilt the data model and visuals; automated data refresh; gathered requirements with the finance team
ImpactCut report creation time from days to minutes and supported executive decisions

Data Cleaning & Automation Script

Internal tool
TechPython (pandas), CSV/Excel
RoleLed development and deployment for internal use
ImpactSaved 15+ hours per month of manual data processing

Cloud Database Project

TechFirebase, Python
RoleDesigned a cloud database and scripts for real-time updates and sync across devices
ImpactApplied data modeling and security rules for efficient, secure storage

Web Data Extraction Project

TechPython, BeautifulSoup, Requests, PostgreSQL
RoleBuilt scraping applications and structured pipelines feeding dashboards
ImpactAutomated pipelines kept downstream reporting continuously up to date

Independent & open-source work

github.com/RaudyG ↗

NBA Game Prediction

Leakage-safe rolling features and a logistic regression model predicting game outcomes.

61.3% accuracy vs. 50% baseline

Real Estate Anomaly Detection

Isolation Forest applied to 526K property records to flag pricing anomalies.

526,000 records

ConsumoElectrico

Electricity-consumption forecasting comparing Linear Regression and Random Forest.

forecasting · regression

World Cup 2026 Predictor

Elo ratings built from scratch plus attack/defense and player-form features; a Dixon-Coles Poisson + Monte Carlo phase is planned next.

in progress · phase 2 planned
Skills

The toolkit.

Languages & core

  • Python
  • SQL
  • Clean coding & automation
  • Algorithm design

ML & data science

  • pandas
  • scikit-learn
  • XGBoost
  • SHAP

Databases & cloud

  • Microsoft SQL Server
  • MySQL / PostgreSQL
  • Firebase
  • Google Cloud (in progress)

Engineering & tools

  • Docker
  • Django REST / Flask
  • Power BI
  • Excel / openpyxl
Education

B.S. Computer Systems Engineering

Catholic and Technological University of Cibao (UCATECI)

01/2019 — 01/2025

Focused on designing tech solutions, project leadership, and process optimization — with an emphasis on efficiency and security in every project.

Certifications

Ongoing training

  • Data Analyst Internship Programme, Uptrail10/2025 — 11/2025
  • Diploma in SQL Server Database Programming
  • Instructional Material Design for the Web
  • Python Certificate
  • GCP Cloud Digital Leader → ML Engineer trackin progress