
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 ...
07/10/2026
Dalet, a leading technology and service provider for media-rich organizations, today announced the latest Long-Term Supported (LTS) release of Dalet Flex. Build...
06/09/2026
June 9 2026, 23:00 (PDT) Dolby and MagentaTV Bring Fans Closer to the FIFA Worl...
04/08/2026
Dalet, a leading technology and service provider for media-rich organizations, t...
22/07/2026
Among other tweaks to coverage, the broadcaster added announce positions in the ...
22/07/2026
The Australian Sports Technologies Network (ASTN) has reported a sharp acceleration in Australia's sportstech sector, with revenue growth rising from 10% in...
22/07/2026
From a retrofitted broadcast bus to packed Match Day Live! events, Bennett detai...
22/07/2026
TMRW Sports has named Brendan Donohue as President, Flag Football, where he will lead the company's professional flag football business in partnership with ...
22/07/2026
Stephen Arnold Music (SAM) designed and produced soundtracks for nine projection...
22/07/2026
Harmonic has announced a $25,000 donation to Global Empowerment Mission (GEM) fo...
22/07/2026
Telemundo has announced it has acquired from UC3 exclusive U.S. Spanish-language...
22/07/2026
The Rest Is Football, hosted by Gary Lineker, Alan Shearer, and Micah Richards, ...
22/07/2026
Sports Studio has announced a partnership with TheSoul Group to distribute Free ...
22/07/2026
Daktronics has concluded its 2026 High School Video Summit, a two-day hybrid event combining virtual sessions with hands-on host school visits across the countr...
22/07/2026
FOR-A America will exhibit at the MMCA Show in Pensacola, Florida (July 21-24), marking the company's first appearance at the house of worship AV event. FOR...
22/07/2026
The New York Yankees have selected Lumen Technologies (NYSE: LUMN) to connect Ya...
22/07/2026
Calrec will exhibit at IBC 2026 (Stand 8.C47) with demonstrations of its IP-native audio ecosystem built around the Argo console range, and will announce two ne...
22/07/2026
Solid State Logic has announced the appointment of Algam EKO as its distributor for SSL large format consoles in Italy, covering the Live, Broadcast, and Studio...
22/07/2026
FloSports has announced the launch of the FloSports Channel, a free ad-supported streaming channel premiering on Prime Video in the U.S. on July 21. The channel...
22/07/2026
Banana Ball TV (BTV), the broadcast operation of the Savannah Bananas, has insta...
22/07/2026
SiriusXM and Audacy have announced an agreement bringing 42 Audacy sports, news, and talk stations from 29 U.S. markets to SiriusXM. Twenty-three Audacy sports ...
22/07/2026
The collaboration produced more than 500 short-form videos across four cities th...
22/07/2026
At 2,600 sq. ft., the new home of digital content houses a main studio, podcast space, and an auxiliary control room
NFL franchises are accelerating their prod...
22/07/2026
Library now available in Engine Player
Another of Eduardo Tarilonte's popular sample libraries has just been moved over to Engine Audio's Engine Pla...
22/07/2026
All-in-one mic & DSP interface overhauled
Cloudvocal have just refreshed the design of their all-in-one mic, audio interface and DSP processor, introducing ...
22/07/2026
The final boss of the lost dual-op-amp circuit
Electro-Harmonix have just introduced the Deluxe Big Muff Pi 2, which they say offers their latest take on t...
22/07/2026
eds3_5_jq(document).ready(function($) { $(#eds_sliderM519).chameleonSlider_2_1({ content_source:......
22/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/07/2026
Proven managed IP solution helps broadcasters confidently navigate the future of broadcast distribution
LTN announces it has completed successful satellite-to-...
22/07/2026
Today, Chaos announced V-Ray 7, update 4, eliminating the need for asset conversions and tool switching, which slow down media and entertainment pipelines. Arti...
22/07/2026
An AI playground for the Project Indigo camera app
Marc Levoy July 22, 2026
0 Comments
At left is the user interface of our Playground, running on an ...
22/07/2026
Custom Consoles announces the completion of a large technical furniture project for Radio T l vision Suisse (RTS) at the networks recently inaugurated productio...
22/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/07/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
22/07/2026
Culver City, CA, July 21, 2026 3D streaming infrastructure provider Miris today announced glTF intake, including the binary .glb form, to its spatial streamin...
22/07/2026
Documentary Lo Invisible Captured with Blackmagic PYXIS 6K
Brie Clayton July 21, 2026
0 Comments
Camera's versatile design allowed filmmakers to c...
22/07/2026
AAF Bridge converts AAF and XML Timelines to Ableton Live - and back
Jerome Gardier July 21, 2026
0 Comments
A two-way converter for macOS and Windows...
22/07/2026
From Class to Concert: Explore a Student-Led Production of David Bowies Work Thr...
22/07/2026
A poisoned spy, a secret war, and a small English city contaminated by a deadly chemical weapon. Available on Sky Documentaries from 4 AugustWednesday 22 July 2...
22/07/2026
Audio at Parity: Comscore Launches Transcript-Level Targeting and Measurement wi...
22/07/2026
Arvato Systems and Lobster Enter into a Partnership and Combine the Integration Platform and Managed Services
Flexible integration solutions for cloud, hybrid...
22/07/2026
RT REPORTS NET SURPLUS OF 22.5 MILLION IN 2025...
22/07/2026
In the Opinion of the Censor airs Monday 27 July at 9.35pm on RT One and RT Player
In the Opinion of the Censor looks at how films for theatrical release wer...
22/07/2026
Before a healthcare robot can be useful in the real world, it has to learn how the physical world pushes back. Anatomy varies. Instruments bend, press, slip and...
21/07/2026
New research involving a cross-section of Mongolias independent media landscape finds enthusiasm for AI as a tool to improve efficiency, sustainability and resi...
21/07/2026
With winners Spain having already paraded the World Cup trophy through Madrid in front of millions of fans and the dust now settling on an epic tournament of th...