Senior data analyst and analytics engineer working with billions of rows of financial transaction data. I build and maintain reliable dbt, SQL, and Python pipelines in BigQuery, design reusable feature layers for analysts and data scientists, and translate technical work into clear stories for non-technical stakeholders.
Pioneered a dbt-based incremental transaction modeling layer for core BigQuery pipelines, cutting bytes billed by 87% and reducing end-to-end runtime by 30%, and integrated it with Airflow to standardize orchestration and monitoring
Built Python tooling in secure Vertex AI Workbench environments to automate complex enrichment workflows, reducing analyst time spent on data enrichment by ~80% for a team of 10+ analysts and enabling faster self-serve analysis
Designed a dbt-driven behavioral profiling layer that transforms transaction data into 50+ reusable user attributes (features), providing a maintainable foundation for profiling, ML, and analytics products
Reworked highly technical presentations on AI-driven merchant identification into clear narratives for engineering and business leadership, tying data pipeline and feature work to company goals and a ~40× increase in identifiable merchants
Analyzed large-scale transactional datasets (10B+ rows) in Vertica using SQL to develop and validate 1,000+ data classification rules for downstream insights and enrichment used by 100+ financial institutions
Built Python automation that reduced core manual processing workflows from 3–4 hours to ~15 minutes, enabling faster iteration and more reliable data delivery
Designed logic for multiple data-driven financial insights deployed to tens of thousands of end users, contributing directly to product value and client-facing analytics
Consolidated data from 15+ projects into a shared PostgreSQL environment by extracting, cleaning, and standardizing datasets for internal research and analytics
Published 50+ company-wide Tableau reports and shaped a data-driven presentation for IT leadership attended by ~85% of the organization
Built a Python system that uses LLMs and web search to turn messy, natural-language place recommendations into structured metadata (address, hours, categories, trip groupings, visit tips). Used Pydantic-AI models to enforce a consistent schema and category set, then generated clean Markdown notes for an Obsidian travel knowledge base.