Tapjoy Uses Amazon SageMaker for Programmatic Mediation in Near Real-Time Advertising

Tapjoy is a mobile advertising and app monetization platform that reaches 1.65 billion active users per month with 40 million daily engagements on mobile apps and devices. Tapjoy's advertising platform places bids in real-time auctions and must handle high throughput with ultra-low latency response times. The company wanted to use machine learning (ML) to predict user behavior and ad pricing for more than 50,000 requests per second, with a requirement to run inferences within a 5-millisecond latency window. Tapjoy also wanted flexibility to test several endpoints in production and keep costs low while minimizing development time for its teams and chose Amazon Web Services (AWS) to build a solution.

Tapjoy used Amazon SageMaker as an end-to-end solution to build, train, and deploy machine learning for near real-time advertising, implementing five single-model and four multi-model endpoints for low-latency inferencing in production. Looking to the future, Tapjoy plans to automate retraining and model deployments, build model monitors, and adopt the SageMaker model registry to track and audit trained models.

“Ultimately, we chose Amazon SageMaker because it has an end-to-end machine learning platform and is very feature rich,” says Nick Reffitt, vice president of data science and engineering at Tapjoy. “We've gone from taking about three to six months to train, build, and deploy a model. Now with SageMaker, we can do that within one week, perhaps even shorter,” says Reffitt. “That has allowed us to move much faster.”

Learn more about advertising intelligence solutions on AWS.