At Intuit, I led a migration of legacy batch pipelines to a lakehouse architecture with Apache Iceberg, AWS S3, and Snowflake. The incremental dbt models and phased zero-downtime rollout cut daily processing time by 40% across billions of rows.
I built a Great Expectations data-contract validation layer to prevent schema drift from reaching production, helping save over $2M annually across three business lines. I also developed a unified semantic layer with LookML and dbt that reduced ad hoc analyst requests by 55%.
At Intuit, I re-engineered credit-risk and default-prediction feature pipelines using PySpark on Databricks. An as-of join framework eliminated feature data leakage and achieved 100% train/serve parity, while feature-job optimization cut extraction runtime 8x.
At Uber, I architected a CDC streaming pipeline using Debezium, Kafka, and PySpark Structured Streaming on Databricks, cutting reporting latency from over 14 hours to under 45 seconds across 180M+ daily events. Earlier, at Centene Corporation, I automated healthcare analytics workflows and built dashboards that eliminated 120+ manual reporting hours per month.
