🌐 Scroll down to keep reading this README in English or Click here for Pt-br version --> Versão em Português
🔍 "Building robust data pipelines and interpretable ML models — from theory to production." 📌 Currently: Data Engineering Intern @ BBTS (Banco do Brasil Tecnologias e Serviços) 🌱 Focus: End-to-end Data Projects (ETL → ML → MLOps) + Backend Development
-
Data Engineering/MLOps: - Python - SQL - ETL - Spark - Azure Databricks - Azure Data Factory - Azure DevOps
-
ML: - Scikit-learn - Feature Engineering - Statisticss - Validation - Azure Databricks
-
Visualization: - Power BI - Matplotlib & Seaborn - Excel - Plotly
-
Backend: - FastAPI - Flask - PostgreSQL - System Design
-
Numerical: - C - Algorithms - Optimization - Numerical Methods
🎓 Student Performance Classifier | Click Here to View Project
- Challenge: Kaggle dataset contained label errors in target variable
- Solution: Manual label correction + SMOTE after train-test split
- Results: 96% Accuracy | F1-score 0.96
- Tech:
PandasSMOTEGridSearchPipeline Design
📉 Churn Prediction (Stats + MLP) | Click Here to View Project
- Approach: Combined hypothesis testing (Chi², Mann-Whitney) with neural networks
- Insights: Age > 40 doubles churn risk | German customers churn more
- Results: 67% Recall | 82% Accuracy
- Tech:
SciPyTensorFlowSHAPStatistical Validation
🏠 Rent Price Prediction | Click Here toView Project
- Dual Approach: Linear regression + price classification (business-interpretable)
- Key Findings: +R$24.03/m² | -R$51.10/km from center
- Results: R² 0.82 | Classification Accuracy 84%
- Tech:
StatsModelsMatplotlibBusiness Rules
-
X-Men Management System Full CRUD for mutant academy with relational DB design (
FastAPISQLAlchemyTkinter UI) -
TaskMaster Collaborative task manager with priority queues (
PythonFlaskBusiness Rules & SQL)
Implemented algorithms from Burden & Faires:
Equation Solving | Numerical Linear Algebra | Eigenvectors and Eigenvalues | Linear and Nonlinear Systems | ODEs
🔗 Repository
- ML with Engineering Rigor: Focus on reproducible pipelines over one-off models
- Cloud-Native Approach: Leverage Azure stack for scalable data solutions
- Production Mindset: From notebooks to deployable systems (learning MLOps)
- Full Pipeline Awareness: From raw data to business insights
📫 Let's connect: LinkedIn | jonashonorato4@gmail.com
🔍 "Construindo pipelines robustos e modelos de ML interpretáveis — da teoria à produção." 📌 Atualmente: Estagiário em Engenharia de Dados @ BBTS (Banco do Brasil Tecnologias e Serviços) 🌱 Foco: Projetos end-to-end (ETL → ML → MLOps) + Desenvolvimento Backend
[](https://holopin.io/@jonasmelo21)
Engenharia de Dados
Python | SQL | ETL | Spark | Databricks | Azure Data Factory
ML/MLOps
Scikit-learn | Engenharia de Features | Validação Estatística | Jupyter Notebooks | Azure Databricks
Visualização de Dados
Power BI | Matplotlib & Seaborn | Excel | Plotly
Backend
FastAPI | Flask | PostgreSQL | Design de Sistemas | Azure DevOps
Computação Numérica
C | Algoritmos | Otimização | Cálculo Numérico | Álgebra Linear Numérica
🎓 Classificador de Performance Estudantil | Ver Projeto
- Desafio: Dataset do Kaggle continha erros nos rótulos da variável alvo
- Solução: Correção manual + SMOTE após divisão treino-teste
- Resultados: 96% Acurácia | F1-score 0.96
- Tecnologias:
PandasSMOTEGridSearchPipelines
📉 Predição de Churn (Estatística + MLP) | Ver Projeto
- Abordagem: Testes de hipótese (Qui², Mann-Whitney) combinados com redes neurais
- Insights: Idade > 40 dobra risco de churn | Clientes alemães saem mais
- Resultados: 67% Recall | 82% Acurácia
- Tecnologias:
SciPyTensorFlowSHAPValidação Estatística
🏠 Predição de Preços de Aluguel | Ver Projeto
- Abordagem Dupla: Regressão linear + classificação de preços (interpretável)
- Coeficientes: +R$24,03/m² | -R$51,10/km do centro
- Resultados: R² 0.82 | Acurácia 84%
- Tecnologias:
StatsModelsMatplotlibRegras de Negócio
-
Sistema X-Men CRUD completo para academia de mutantes com banco relacional (
FastAPISQLAlchemyTkinter UI) -
TaskMaster Gerenciador de tarefas colaborativo com filas de prioridade (
PythonFlaskRegras de Negócio)
Algoritmos implementados do livro Burden & Faires:
Solução de Equações | Álgebra Linear Numérica | Autovetores e Autovalores | Sistemas Lineares e Não Lineares | EDOs
🔗 Repositório
- Rigor em Engenharia de ML: Pipelines reprodutíveis em vez de modelos isolados
- Abordagem em Nuvem: Uso do stack Azure para soluções escaláveis
- Mentalidade de Produção: Do notebook à implantação (aprendendo MLOps)
- Visão do Pipeline Completo: Dados brutos até insights de negócio
📫 Vamos conversar: LinkedIn | jonashonorato4@gmail.com
