As a Ph.D. Candidate in applied machine learning at the Knowledge Discovery Lab in UMB’s CS dept, my focus is on modeling real-world datasets with deep neural networks in Python. Research projects have included: modeling human activity recognition with wearable accelerometers (from perspectives including: labeling uncertainty, ablation studies with respect to number of sensors / handedness of individuals, effects of subject-specific distributional differences on modeling), fall risk modeling in older adults, crime forecasting, music genre recognition, and stream flow prediction for river flooding. Supervised undergrad lab assistants and mentored more junior Ph.D. students in data science methodology and programming best practice.

Worked with Indigo’s data science team to design, implement, train, and evaluate an irrigation status model for cropland in the continental US, trained on multispectral satellite imagery (to 30m resolution) and associated weather data. Used Python, AWS, SQL, and XGBoost. Results published as an abstract at AGU 2019: https://bit.ly/2xsPyvj.

Assisted the decision-making of corporate and underwriting staff by providing statistical risk metrics quantifying the company’s risk exposure to hurricanes and earthquakes. Responsible (with team) for populating and maintaining data store (MS SQL) of direct property building data, for use in risk modeling and accumulation management. Designed, developed, and maintained internal tools to automate catastrophe department workflow and reporting using C# and SQL. Identified and remedied inefficiencies in existing analyst workflows and screened future changes to procedures for potential issues. Suggested alternatives and coded solutions. Was personally responsible for the intake and processing of another multinational property insurer’s in force policy data following a major acquisition. Moved up in the company several times, finishing as AVP – Catastrophe Modeling (2014-2016). Worked as a Senior Catastrophe Analyst from 2010-2014 and as Catastrophe Analyst from 2005-2010.