
Pinterest has engineered a way to serve its photo-sharing community more of the images they love.
The social-image service, with more than 400 million monthly active users, has trained bigger recommender models for improved accuracy at predicting people's interests.
Pinterest handles hundreds of millions of user requests an hour on any given day. And it must also narrow down relevant images from roughly 300 billion images on the site to roughly 50 for each person.
The last step - ranking the most relevant and engaging content for everyone using Pinterest - required a leap in acceleration to run heftier models, with minimal latency, for better predictions.
Pinterest has improved the accuracy of its recommender models powering people's home feeds and other areas, increasing engagement by as much as 16%.
The leap was enabled by switching from CPUs to NVIDIA GPUs, which could easily be applied next to other areas, including advertising images, according to Pinterest.
Normally we would be happy with a 2% increase, and 16% is just a beginning for home feeds. We see additional gains - it opens a lot of doors for opportunities, said Pong Eksombatchai, a software engineer at Pinterest.
Transformer models capable of better predictions are shaking up industries from retail to entertainment and advertising. But their leaps in performance gains of the past few years have come with a need to serve models that are some 100x bigger as their number of model parameters and computations skyrockets.
Huge Inference Gains, Same Infrastructure Cost Like many, Pinterest engineers wanted to tap into state-of-the-art recommender models to increase engagement. But serving these massive models on CPUs presented a 100x increase in cost and latency. That wasn't going to maintain its magical user experience - fresh and more appealing images - occurring within a fraction of a second.
If that latency happened, then obviously our users wouldn't like that very much because they would have to wait forever, said Eksombatchai. We are pretty close to the limit of what we can do on CPU basically.
The challenge was to serve these hundredfold larger recommender models within the same cost and latency constraints.
Working with NVIDIA, Pinterest engineers began architectural changes to optimize their inference pipeline and recommender models to enable the transition from CPU to GPU cloud instances. The technology transition began late last year and required major changes to how the company manages workloads. The result is a 100x gain in inference efficiency on the same IT budget, meeting their goals.
We are starting to use really, really big models now. And that is where the GPU comes in - to help make these models possible, Eksombatchai said.
Tapping Into cuCollections Switching from CPUs to GPUs required rethinking its inference systems architecture. Among other issues, engineers had to change how they send workloads to their inference servers. Fortunately, there are tools to assist in making the transition easier.
The Pinterest inference server built for CPUs had to be altered because it was set up to send smaller batch sizes to its servers. GPUs can handle much larger workloads, so it's necessary to set up larger batch requests to increase efficiency.
One area where this comes into play is with its embedding table lookup module. Embedding tables are used to track interactions between various context-specific features and interests of user profiles. They can track where you navigate, and what people Pin on Pinterest, share or numerous other actions, helping refine predictions on what users might like to click on next.
They are used to incrementally learn user preference based on context in order to make better content recommendations to those using Pinterest. Its embedding table lookup module required two computation steps repeated hundreds of times because of the number of features tracked.
Pinterest engineers greatly reduced this number of operations using a GPU-accelerated concurrent hash table from NVIDIA cuCollections. And they set up a custom consolidated embedding lookup module so they could merge requests into a single lookup. Better results were seen immediately.
Using cuCollections helped us to remove bottlenecks, said Eksombatchai.
Enlisting CUDA Graphs Pinterest relied on CUDA Graphs to eliminate what was remaining of the small batch operations, further optimizing its inference models.
CUDA Graphs helps reduce the CPU interactions when launching on GPUs. They're designed to enable workloads to be defined as graphs rather than single operations. They provide a mechanism to launch multiple GPU operations through a single CPU operation, reducing CPU overheads.
Pinterest enlisted CUDA Graphs to represent the model inference process as a static graph of operation instead of as those individually scheduled. This enabled the computation to be handled as a single unit without any kernel launching overhead.
The company now supports CUDA Graph as a new backend of its model server. When a model is first loaded, the model server runs the model inference once to build the graph instance. This graph can then be run repeatedly in inference to show content on its app or site.
Implementing CUDA Graphs helped Pinterest to significantly reduce inference latency of its recommender models, according to its engineers.
GPUs have enabled Pinterest to do something that was impossible with CPUs on the same budget, and by doing this they can make changes that have a direct impact on various business metrics.
Learn about Pinterest's GPU-driven inference and optimizations at its GTC session, Serving 100x Bigger Recommender Models, and in the Pinterest Engineering blog.
Register for GTC, running Sept. 19-22, for free to attend sessions with NVIDIA and dozens of industry leaders.
Most recent headlines
05/01/2027
Worlds first 802.15.4ab-UWB chip verified by Calterah and Rohde & Schwarz to be ...
01/06/2026
January 6 2026, 05:30 (PST) Dolby Sets the New Standard for Premium Entertainment at CES 2026
Throughout the week, Dolby brings to life the latest innovatio...
02/05/2026
Dalet, a leading technology and service provider for media-rich organizations, t...
01/05/2026
January 5 2026, 18:30 (PST) NBCUniversal's Peacock to Be First Streamer to ...
01/04/2026
January 4 2026, 18:00 (PST) DOLBY AND DOUYIN EMPOWER THE NEXT GENERATON OF CREATORS WITH DOLBY VISION
Douyin Users Can Now Create And Share Videos With Stun...
23/03/2026
The Professional Fighters League (PFL) has renewed its multi-year partnership wi...
23/03/2026
The Snow League has named Google Cloud as its Official Cloud and AI Partner. The...
23/03/2026
Chyron has appointed Eric Wolff as Director of Venues Sales, North America. Wolff previously served as Director of Broadcast Operations & Media Production for T...
23/03/2026
Chicago Sports Network (CHSN) and Weigel Broadcasting's WCIU (The U, ch. 26.1) will simulcast 10 Chicago White Sox games during the 2026 season, the compani...
23/03/2026
Cosm has appointed Jon Werbeck as Vice President, Head of Sponsorships. He will report to Corey Breton, Head of Venues, and will focus on corporate sponsorship ...
23/03/2026
CP Communications has announced a partnership with Mark Roberts Motion Control (...
23/03/2026
NAB Show 2026, taking place April 18-22 (exhibits April 19-22) at the Las Vegas ...
23/03/2026
Bay FC and free streaming platform Victory have announced a partnership through...
23/03/2026
Gemini AI models will surface hidden context around pitches, matchups, rare stat...
23/03/2026
Behind The Mic provides a roundup of recent news regarding on-air talent, includ...
23/03/2026
Growing from broadcast engineer to strategic planner, this Ithaca College grad h...
23/03/2026
16 Science-Focused Nonfiction Projects Selected for Funding
LOS ANGELES, CA, March 23, 2026 - The nonprofit Sundance Institute and Sandbox Films announced toda...
23/03/2026
It's been 20 years since Miley Cyrus introduced the world to Hannah Montana,...
23/03/2026
Made entirely from real natural recordings
Aimed at sound designers and editors working in film, TV and game audio, the latest release from BOOM Library com...
23/03/2026
Transcribe sheets, tabs or MIDI from audio files
Klang.io have announced the launch of a new AI-powered software tool that's capable of detecting multip...
23/03/2026
An auxiliary target has been affixed to the Interim Cryogenic Propulsion Stage f...
23/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/03/2026
Pro8mm, the Super 8 experts, provided cameras, Super 8 movie film, and scanning services for Bruno Mars' Risk It All music video. The debut single from Br...
23/03/2026
Matthews, introduces their first aluminum grid clamp collection, engineered for the rigging needs of film, television and live production. Combining light weigh...
23/03/2026
Monday 23 March 2026
Hacks, the multi-Emmy -winning Sky Exclusive comedy, retur...
23/03/2026
Back to All News
Too Hot to Handle: Italy Reignites for a Second Season With th...
23/03/2026
Autonomous agents mark a new inflection point in AI. Systems are no longer limited to generating responses or reasoning through tasks. They can take action: Age...
23/03/2026
RT is sad today to learn of the death of legendary RT Sport broadcaster Michael Lyster, who died this morning aged 71 years.
Kevin Bakhurst, Director-General...
23/03/2026
RT Documentary On One has scooped its first ever dedicated music award. At the 2026 Icelandic Music Awards, composer lfur Eldj rn won Release of the Year in t...
23/03/2026
Inside Sport, Liveline, Morning Ireland and 2FM DRIVE will all be in Prague to bring fans to the heart of the action
Every Moment, Every GenerationRT | FIFA W...
22/03/2026
Free updates now available
VSL have just released some free updates that add some existing features to a selection of libraries in their expansive Synchron ...
22/03/2026
Back to All News
Live-Action Sins of Kujo' Premieres April 2: Main Trailer and Key Art Debut
Entertainment
22 March 2026
GlobalJapan
Link copied to cl...
21/03/2026
Presented to War Child UK's HELP(2) project
The MPG (Music Producers Guild) have announced the launch of the MPG Impact Award, a brand-new honour that w...
21/03/2026
Microtuning support for Arabic, Persian & Turkish scales
The latest release from Best Service brings together a selection of string, wind and percussion ins...
21/03/2026
New campaign from NAATI and SBS CulturalConnect highlights how we all deserve t...
21/03/2026
Statement regarding Rhoda Roberts AO
21 March, 2026
Media releases
SBS is deeply saddened by the passing of Widjabul Wia-bal woman from the Bundjalung Na...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
21/03/2026
Cine Gear Connect NY, presented by Universal Production Services, is filling in the slate for a full day of panels, peers, learning the latest, and mixing it up...
21/03/2026
Studio Technologies Debuts New StudioComm System at NAB 2026
Brie Clayton March 20, 2026
0 Comments
StudioComm Model 794 Central Controller and Model ...
21/03/2026
Restoration Christian Fellowship Captures Worship Music Videos with PYXIS 12K
Brie Clayton March 20, 2026
0 Comments
PYXIS' open gate provides cre...
20/03/2026
Net Insight will introduce a JPEG XS solution for full IP environments at NAB Sh...
20/03/2026
LTN has expanded its technology partnership with Harmonic ahead of the FCC's...
20/03/2026
Solid State Logic will preview SSL Live V6.2 at NAB Show, booth C6907. The softw...
20/03/2026
FUJIFILM North America Corporation's Optical Devices Division has announced ...