Professional Documents
Culture Documents
Medellín, Colombia
Phone number: +57 3233804664 E-mail: brahyanmaiko@gmail.com
Linkedin: [ linkedin.com/in/brahyan-jimenez-aricapa-ba0529189 ]
GitHub: [ https://github.com/Brahyanmaiko ]
Work Experience
● ML Risk back-testing models Analyze, monitor, retrain current models for risk management team,
models deployed on azure databricks.
● Perform financial data analysis to generate reports that help the company make informed decisions
regarding credit risk.
● Develop financial and risk models to aid in the company's decision-making process.
● Implement data analysis tools and techniques to identify trends, patterns, and improvement
opportunities in credit risk management.
● Collaborate with other departments to ensure accuracy and integrity of financial data used in reports.
● Communicate findings and recommendations to leadership team members to help them make
informed decisions about credit risk.
● Contribute to the development of credit risk management policies and procedures to ensure
compliance with industry regulations and requirements.
Main Technologies: Python, SQL, Pyspark, Pandas, Scikit-learn, Azure Databricks, Azure Data Factory, Power BI,
Excel.
Projects
● Predicted whether a request or connection to a server is a cyber-attack. Collected the data from different
sources via SQL queries, initial dataset size was +90.000 samples and 152 features, pre-processed it,
created new features compressed using Autoencoders approach, increasing the number of features to
172, analyzed correlations between the all the features regarding the target with statistical methods,
leaving the top 20 features with more correlation score. Trained different Machine Learning models and
looking for the best results with the lower number of features to reduce the complexity and time
inference of the model, getting as final model a GradientBoostingClassifier using only 2 features
achieving 99.94% accuracy.
Main Technologies: SQL, python, pandas, numpy, Scikit-learn, Keras
● Predicted the desertion of a student based on several factors such as Ethnicity, Gender, Scores, UCAS
Points among others. Cleaned, Imputed and Applied Feature Engineered to Data, Analyzed the current
data looking for high correlations between features, eliminated those with high correlation with SULOV
method using a XGBoost Machine Learning Model, then Balanced the data regarding the target creating
artificial samples using SMOTE. Selected evaluation metrics and baseline models. Trained and tested
different models, best model found was a RandomForestClassifier achieving 0.97 accuracy, 0.93 AUC,
0.88 recall, 0.95 precision, 0.91 F1 score.
Main Technologies: Python, pandas, numpy, Scikit-learn, XGBoost
Skills
Tech Skills: Python, C++, SQL, MySQL, Numpy, Pandas, Keras, Scikit-learn, Pytorch, Docker, AWS.
Agile Methodologies: Scrum
Other Tools: IOT, ESP32, Sensors, MATLAB.
Languages: Fluent in English, Spanish Native
Education