I'm a Data Analyst Intern at Banco de Brasília (BRB), where I focus on Machine Learning model monitoring, validation, data automation, and reporting for risk-related models. My background in Physics gave me a strong foundation in mathematics, statistics, and analytical thinking, which naturally led me toward data science and machine learning.
I'm currently exploring new opportunities in Data Analytics and Data Science, with a focus on remote roles.
- Automated end-to-end ML model monitoring across 20+ Credit Risk model segments, Liquidity Risk (6 models), and IPCA — reducing processing time by over 95% through Python automation and a custom GUI tool.
- Led the development of a multi-agent AI architecture for preliminary model validation assessment, orchestrating 5+ specialized agents on Gemini Enterprise.
- Design and redesign strategic Power BI dashboards using Power Query (M) and DAX for large-scale data transformation and reporting.
- I write about data science, machine learning, SQL, and analytics on Medium.
Selected initiatives from my experience at Banco de Brasília (BRB). Details are intentionally presented at a high level due to the proprietary nature of internal systems and data.
| Initiative | Description |
|---|---|
| ML Model Monitoring Automation | Automated end-to-end monitoring workflows across 20+ Credit Risk model segments, 6 Liquidity Risk models, and IPCA using Python, Hive, APIs, and a custom GUI tool. Developed automated chart generation and scripts to embed analytical outputs directly into Word reports, reducing processing time by over 95%. |
| AI-Assisted Model Validation | Led the development of a 6-agent architecture for preliminary model validation assessment on Gemini Enterprise. The workflow includes agents for document eligibility assessment and solution classification (model vs. methodology), 3 independent specialist agents for technical assessment, and a final agent responsible for generating the Technical Assessment Report (TAP). |
| Risk Analytics & Power BI | Built and redesigned 3 strategic Power BI dashboards covering Liquidity Risk, RWACPAD, and IPCA (12-month and monthly views). Used Power Query (M) and DAX to support model monitoring, risk analysis, and reporting, while designing the dashboards for efficient recurring updates and maintenance. |
|
Languages & Data Analysis
|
Machine Learning
|
BI & Data Visualization
|
Tools & Infrastructure
|
| Project | Description | Links |
|---|---|---|
| ORMS — Operating Room Management System | Relational database designed to manage hospital operating room workflows, including patients, procedures, surgeries, medical staff, equipment, and scheduling resources. Implemented normalization, foreign keys, constraints, indexes, analytical views, and SQLite triggers to enforce data integrity and prevent scheduling conflicts. | Code · Article |
| CEAPS Data Analysis & Forecasting | End-to-end analysis of 386K+ Brazilian Senate expense records (2008–2026); benchmarked SARIMA, ARIMA, Linear Regression, and Prophet for expense forecasting (SARIMA, 10.5% MAPE). | Code · Article |
| A/B Testing — E-commerce Recommendation | Full A/B testing pipeline on 294K+ observations to evaluate a product recommendation system's impact on conversion rate, using one- and two-tailed Z-tests for proportions. | Code · Article |
| Movie Recommendation API (MovieLens) | KNN-based recommendation system (32M+ ratings) deployed as a production-style REST API with FastAPI, Pydantic validation, and Docker containerization. | Code · Article |
- Phishing Website Detection (ML Classification) — Classification model to detect phishing websites using URL-based features. Applied EDA, feature engineering, data preprocessing, and classification models, with evaluation based on accuracy, precision, recall, and F1-score. Currently being reworked and improved from an earlier version. → Repo
- AWS — Cloud Practitioner Essentials
- SQL — freeCodeCamp SQL course
- From Grey Sloan's Whiteboard to a Relational Database
- Recommendation Engines: How to Build One with MovieLens
- I Tested a Product Recommendation System. The Result Was "No" — and That's the Point
- O que os dados revelam sobre os gastos do Senado?
More articles on Medium.
