Natali Kostadinovic

I am a statistics student and data scientist experienced in preprocessing complex biomedical signals, engineering features, and building predictive models that improve accuracy and reduce manual effort. I have applied techniques including logistic regression, XGBoost, cross-validation, and SHAP to achieve strong predictive performance.

In research, I built a preprocessing pipeline for 128-channel EEG (filtering, artifact rejection, ICA) that boosted model accuracy to 85% and reduced manual cleaning time by 20%, and engineered ERP and acoustic features to improve perception modeling. I collaborate with faculty and peers in Google Colab to test methods and streamline workflows.

I also bring practical industry QA experience testing app features with TestRail and Postman, designing test cases that cut post-release defects and mentoring offshore QA teams. I enjoy turning data into actionable insights and building dashboards, forecasts, and reproducible pipelines.