Graph Data Scientist- Network analysis for production data- Machine learning for graph-structured data- Simulations of production networks

Extensive experience generating new insights from different types of data based on robust statistical methods, machine learning and innovative approaches:- Public sector: Led development of novel approaches for Anomaly Detection using Graph/Network algorithms on a multi-terabyte SQL database at a federal institution. Created pipeline for graph creation and ranking of connected entities (PageRank, label propagation, network metrics) as well as applying machine learning on graphs (node classification);- Public sector: Developed outlier and anomaly detection models for ship GPS position and movement data streams;- Healthcare: Analyzed electronic health record (EHR) data from a large patient database. Extracted temporal features from hospital admissions, doctor visits, laboratory test values, and demographics. Trained a machine learning model to predict the probability of hospitalization in a cohort of 110k SARS-CoV-2-positive patients. Coordinated activities in project team of eight. Publication for the American Medical Informatics Association (AMIA) Symposium 2021 (http://arxiv-export-lb.library.cornell.edu/abs/2110.15002v1);- Automotive: Developed a model to predict vehicle performance from sensor, weather and geographic data, using XGBoost;- Banking: Data clustering, EDA and quality assessment of an international bank’s customer database.

Roche Diabetes Care Germany- Developed and evaluated business cases for go-to-market projects for new technologies in the management of Diabetes

Strategic projects in Germany and Switzerland, including:- Global Business Development, Diagnostics M&A: screening of biomarker diagnostics companies with new proprietary content according to scientific criteria and financial potential.- Marketing and Product Management, Hematology- Project Office for the General Manager, Germany

Screening of biomarker diagnostics companies with new proprietary content according to scientific criteria and financial potential.

Roche Advanced Analytics Data Challenge, 9th place Roche-wide (out of 132 participating teams), 12/2017–01/2018- Team developed predictive model for 1-year mortality in cancer patients from demographic, biomarker, claims and diagnostic test data. Best model was ensemble of Boosted Trees (XGBoost) and logistic regression on engineered features (date information, Latent-Dirichlet Allocation, ICD codes).Kaggle Competitions (selection of solo participation) Since 2017- Personalized Medicine: Redefining Cancer Treatment (Top 11%, place 140/1386): neural network and XGBoost ensemble- Mercedes-Benz Greener Manufacturing (Top 16%, place 596/3835): neural network for regression on engineered features

Studied the role of AMPA and NMDA receptors in the formation and retrieval of emotional memories using genetically engineered mice.

Analyzed the gene expression changes in the mouse brain that occur during the extinction of fear memories, as well as the role of epigenetics in this process.