Pinterest's Trust-Focused AI Discovery Engine and AWS Infrastructure

Pinterest runs its AI discovery engine, which processes over 500 petabytes of data and serves 600 million users per month, on Amazon Web Services. This system, prioritizing trust and user inspiration, has enabled the company to record significant revenue growth.

AI and Infrastructure Capacity

Pinterest utilizes a robust infrastructure to serve 600 million monthly active users and recommend the right image in real-time from among billions of options. The platform has been operating in collaboration with Amazon Web Services since 2010.

Real-Time Recommendations with AWS

The system includes Amazon Elastic Kubernetes Service and more than 10,000 EC2 G5 compute instances used for live recommendations. Additionally, more than 600 customized instances training 18 terabytes of data daily operate within this architecture.

Processing Volume in Millions Per Second

Thanks to this comprehensive technological infrastructure developed, the platform is able to offer more than 10 million AI-powered recommendations to users every second and instantly process large datasets.

Visual Search and Advertising Projects

While projects like Pinterest Canvas generate high-resolution visuals for advertisers, visual search tools powered by Amazon SageMaker successfully recognize more than 2.5 billion objects in photos.

Voice Search and Growth Rates

While the newly introduced voice AI assistant allows users to search and create plans, approximately 70 percent of content discovery is driven by artificial intelligence. These developments provided a 17 percent increase in annual revenue.

Security and Chief Architect's View

Chief Architect Kartik Paramasivam emphasized that the element of trust continues to be the most fundamental building block of the platform for filtering inappropriate content and personalized searches.