I've built end-to-end machine learning pipelines at DrivenData and DKFZ/EMBL, automating predictive modeling and variant scoring across 5,000+ datasets while making model decisions visible through SHAP.
During my doctoral research with the ATLAS Experiment at the University of Siegen and CERN, I trained deep neural-network classifiers on more than 10 million high-dimensional records, achieving 79% signal/background classification accuracy. I owned statistical analysis pipelines from raw detector data to peer-reviewed results, applying Bayesian inference, uncertainty quantification, and reproducible data-quality methods.
