PortfolioRafael Caetité

Software and data engineer.

Minas Gerais: Brazil

Available for new opportunities

Scroll

Development Process

Scroll to explore

I handle the complete data lifecycle: from building ingestion pipelines and training models to delivering functional applications.

01Data
02Model
03Delivery

Data Engineering & Processing

Building automated pipelines to extract, clean, and structure large datasets, ensuring data quality and reproducibility.

Python & SQL · ETL/ELT · Apache Spark · Airflow · dbt

Machine Learning & Analysis

Developing and validating predictive models with strict out-of-time evaluation to guarantee robustness in production.

Machine Learning · Scikit-Learn · PyTorch · MLOps

APIs & User Interfaces

Deploying models as scalable APIs and building responsive frontends so end-users can easily interact with the data.

FastAPI & REST APIs · Docker · Cloud (AWS/GCP) · CI/CD

Stage

Selected work.

PluvioMT

Climate ML · IC/TCC research continuation

Machine learning applied to forecasting rainy season indices in Mato Grosso.

A Medallion data pipeline on Databricks (PySpark, Delta Lake) processing 25 years of climate data to evaluate rainfall onset and demise via LSTM and XGBoost with Walk-Forward validation. The predictive engine feeds an interactive interface that presents double-cropping viability to maximize producers' yields.

2000—2025
Series: 5
Sources: 2 / 3
Targets / Definitions

Python
PySpark
Delta Lake
Databricks
XGBoost
LSTM
React

Breast Cancer Mortality

Data Science · UFV INF493

Predictive mortality model based exclusively on the patient's initial clinical and genetic profile.

Developed as a Data Science final project utilizing LightGBM over the METABRIC dataset with 600+ genetic features. The model isolates post-diagnosis treatments for causal rigor, and uses SHAP explainability to reveal exactly which mutations indicate the highest risk of imminent mortality.

1.9k
Patients: 600+
Features: 0.68
AUC-ROC

Python
LightGBM
scikit-learn
Pandas
METABRIC

inicio

Árvores Cadastradas

████

Vistorias Realizadas

██

Laudos Gemini AI Emitidos

███

Matriz de Risco Operacional

Alerta Laranja: Rajadas > 40km/h

Previsão 24h

Atividades Recentes

Arbolia

SaaS · Geospatial

Georeferenced urban forestry management platform for offline-first tree inventorying and technical ISA risk assessments.

An enterprise PWA for field engineers featuring extreme offline resilience via IndexedDB and Firebase NoSQL. It stands out with purely client-side vector PDF generation and a fail-safe Gemini AI pipeline that translates complex botanical risk matrices into highly accessible technical reports.

Offline-first
Architecture: Gemini
AI Integration: Leaflet
Geospatial

React
TypeScript
Firebase
Firestore
Leaflet
Gemini API
Tailwind

RefManager ProWorkspace: Thesis_Final_v2

Web Worker: Active

Local Processing

Collections

All References978

Deduplicated854

Smart Folders

Machine Learning

Climate Models

Systematic Reviews

Similarity Threshold:

95%

Year

Title & Authors

Source

Actions

2023

Deep learning for climate modeling: A systematic reviewSmith, J., Doe, A.

Scopus

2022

Attention is all you need in time series forecasting

Merged 3 duplicates

Vaswani, A., et al.

Web of Sci

2024

Machine learning applications in precision agricultureJohnson, M., Williams, K.

PubMed

Processing references...Running Levenshtein distance on 12,450 title pairs

45%Worker Thread

RefManager Pro

Data Processing

High-performance client-side engine for intelligent consolidation and deduplication of massive bibliographic datasets.

Built to process thousands of records across disparate sources (Scopus, WoS, IEEE). Features resilient CSV parsing, Regex auto-detection, and optimized Gestalt Pattern Matching running in a parallel Web Worker thread to prevent UI blocking. Concludes with intelligent LLM batching logic for AI ingestion, all securely processed 100% in-browser.

Gestalt LCS
Algorithm: Web Worker
Concurrency: 100% Local
Privacy

React
TypeScript
Tailwind
Web Workers
LLM Batching

flavia.com.br

Custom Web Applications

Freelance · Talon Tech

End-to-end web applications focused on solving operational requirements for commercial clients.

Development of custom solutions for clients without internal engineering teams, encompassing the entire product lifecycle. Delivered scheduling systems and administrative panels that streamline operations, reduce manual bottlenecks, and support customer acquisition.

15+
Endpoints: 2023
Since

React
Vite
FastAPI
PostgreSQL
MySQL
Leaflet

Experience.

20252026
Viçosa, MG
Undergraduate Research Fellow — PIBIC/CNPq
Universidade Federal de Viçosa
Applied research in machine learning and climate data, encompassing data ingestion, predictive modeling, and application development.
- Production of two literature reviews on Retrieval-Augmented Generation (RAG) and Machine Learning applied to rainy season forecasting.
- Implementation of a complete forecasting pipeline with automated data ingestion from Google Earth Engine, utilising CHIRPS, ERA5, Sentinel, and MODIS.
- Training of machine learning models to forecast the end of the rainy season and deterministically predict its duration.
- Development of an agroclimatic dashboard using React and Vite for interactive visualization of forecasts.
- Python
- Machine Learning
- RAG
- Data Pipelines
- Google Earth Engine
2023Present
Minas Gerais
Freelance Developer
Talon Tech
Full-stack development focused on building internal systems and optimizing workflows for external clients.
- Development of 15+ REST endpoints in Python and FastAPI supporting client web applications.
- Design and optimization of PostgreSQL and MySQL schemas, reducing query times through indexing and normalisation.
- Development of responsive React front ends integrated with backend services.
- Hardware and software troubleshooting, focusing on the provisioning, configuration, and maintenance of client workstations.
- Python
- FastAPI
- PostgreSQL
- RESTful APIs
- React
- Full-Stack
05/202201/2028
Viçosa, MG
BSc Computer Science
Universidade Federal de Viçosa
Comprehensive bachelor's program with a strong foundation in mathematics, algorithms, and software engineering at one of the top 10 universities in Brazil.
- Algorithms and data structures, databases, computer networks, and statistics.
- Data Structures & Algorithms
- Relational Databases
- Software Engineering
- Statistics

Stack.

01Languages: PythonSQLJavaScriptTypeScriptJavaC++C#
02Data & modelling: PandasNumPyscikit-learnXGBoostRandom ForestMLP
03Geospatial: NetCDFCopernicus CDSGoogle Earth EngineRasterio
04Services: FastAPIRESTSpringDocker
05Interface: ReactViteNext.jsTailwindAngularHTML/CSS
06Storage & tooling: PostgreSQLMySQLGitLinuxLaTeX

Rafael Caetité — Organizing Director of the 1st Scientific Exhibition of UFV's Computer Science Week (Tech & Science).

Computer Science undergraduate at UFV (Jan 2028), focused on building Data and Artificial Intelligence solutions.

Fascinated by computers since childhood, I started out troubleshooting and repairing hardware, which naturally evolved into a passion for solving complex problems through programming.

During my research fellowship, I also established myself as a researcher, publishing academic articles on machine learning applied to climate sciences. In practice, I built the pipelines to process decades of satellite data, trained predictive models, and developed dashboards designed to help farmers in Mato Grosso maximize their crop production.

Outside of research, I've been working as a freelance developer since 2023. I build custom websites and web apps for clients, focusing on API integrations and relational database design.

Based in: Minas GeraisBrazil
Studying: BSc Computer Science, UFV
Languages: Portuguese (Native), English (Fluent), Spanish (Advanced).

Get in touch.

Resume