Pinterest has built a new AI layer with NVIDIA to support a growing range of products that rely on both images and language, including visual search, content understanding and safety systems.
Built on the NVIDIA accelerated computing platform with NVIDIA Blackwell GPUs , NVIDIA Dynamo and Pinterest’s own visual embeddings, the new foundation gives Pinterest teams a shared way to build and launch multimodal AI products without creating custom infrastructure for each one.
With NVIDIA, Pinterest is turning 80 billion+ monthly searches into high-value signals for AI-powered discovery and shopping, enabling Pinterest Assistant to run 25x more visual context per request. In benchmark testing, Pinterest saw average improvements of about 85x faster response startup and 7.3x faster overall latency when using precomputed visual representations instead of repeatedly processing raw images.
“Building the next generation of AI-powered discovery means investing in infrastructure that can keep up with the scale and complexity of Pinterest,” said Kartik Paramasivam, Chief Architect at Pinterest. “Our collaboration with NVIDIA helps us deliver faster, smarter and more personalized experiences for the hundreds of millions of people who use Pinterest.”
Built for the Pinterest Intelligence and beyond
Pinterest’s foundational system was initially developed to support our intelligence layer, the technology woven throughout the user experience. For multimodal serving, Pinterest Intelligence combines NVIDIA Blackwell GPUs and NVIDIA Dynamo with open-source models and Pinterest-built technology, alongside NVIDIA’s broader AI software stack.
This foundation already supports a broader set of AI use cases across our platform, including ranking, safety, OCR, signal generation and future experiences. As we expand our use of vision-language models (VLMs) for features such as Pinterest Assistant, we need stronger infrastructure to make these experiences faster and more efficient at scale. Our expanded partnership with NVIDIA allows us to do exactly that.
“Pinterest is transforming visual discovery with AI, helping hundreds of millions of people find inspiration through more intelligent and personalized experiences,” said Ujval Kapasi, vice president, AI & HPC Frameworks and Libraries at NVIDIA. “By building its multimodal AI infrastructure on NVIDIA accelerated computing and software, Pinterest can bring the next generation of visual and conversational AI experiences to its users at scale.”
A shared platform for what comes next
This work builds on nearly five years of collaboration between Pinterest and NVIDIA across over 14,000 NVIDIA GPUs. By standardizing the infrastructure behind multimodal AI, Pinterest is giving teams a shared foundation for building new experiences while improving performance and efficiency behind the scenes.
Instead of a single product launch, the end result is a long-term platform for the next generation of visual and conversational experiences across Pinterest.
For more information on Pinterest’s collaboration with NVIDIA, check out our engineering blog.
