At Metropolitan Family Services, I built a Zoom-to-BigQuery pipeline that transcribes weekly outreach calls and extracts structured incident records, cutting manual note-taking by three hours each week.
I modeled a 120GB BigQuery warehouse in dbt Core, bringing Apricot, REDCap, and Salesforce data into one governed source behind eight Looker dashboards. This reduced reporting discrepancies by 30%.
I also engineered event-driven ingestion and data quality checks that load 50,000 records weekly, while an LLM schema mapper reduced manual intervention when partner formats changed. KPI-driven models helped program leadership direct outreach and violence-intervention staff toward high-risk areas, contributing to a 15% reduction in violent incident hotspots.
Previously, at Data Edge, I used SQL and Tableau to analyze multi-client sales data, helping reduce overstock by 18% and manual reporting effort by 35%. In personal projects, I built an XGBoost model for car insurance risk segmentation and a RAG assistant for answering questions about car ownership documents.
