Case Study: GumGum achieves 5x faster data pipelines and 2x lower infrastructure costs with Databricks

Contextual digital advertising at scale with data and analytics

GumGum, a contextual advertising platform, needed real-time bidstream analytics to deliver relevant ads at scale—processing more than 35 billion events per day across 100+ pipelines. Their legacy stack (AWS EMR and Apache Storm) struggled with autoscaling, spot-instance management, fast analytics, and cross-team collaboration, limiting model deployment and timely business insights.

By moving to Databricks with Spark Streaming and Delta Lake, and adopting MLflow, Looker integration, and interactive notebooks, GumGum unified data engineering, data science, and analytics. The platform accelerated pipelines 5x, halved infrastructure costs, achieved ad-forecast response times under 30 seconds, improved model deployment and team collaboration, and enabled cost-effective autoscaling and spot-instance use.

GumGum

Rashmina Menon

Senior Data Engineer