San Francisco, California
Re-architected the Ad to lead attribution pipeline on a richer dataset of user-business interactions, resulting in a 30% lift in ad attributed leads and a 4x increase in pipeline performance.
Re-architected core daily ad campaign snapshot metrics calculating budget proration, daily ad spend, estimated and actual spend pacing, and maximum billable spend, resulting in a 2x increase in pipeline performance and greatly increased visibility in overall ad campaign performance.
Migrated Ads recall system from HAProxy to an Envoy mesh and moving commonly interacting services to the same geographical cluster, greatly reducing P95 to P99 latencies.
Improved distributed ad impression to click attribution Spark ETL pipeline by using pre-computed metrics from historical runs, resulting in a 2x increase in performance and 2x reduction in operational costs.