Condé Nast moderates 50,000+ user submissions with Amazon Nova Pro

Digital audiences increasingly expect to engage directly with the brands they love—not just consume content passively. For Condé Nast, the global media company behind Vogue, Vanity Fair, Glamour, GQ, Condé Nast Traveler/Traveller, and Bon Appétit, this meant creating safe, interactive spaces that could spark conversations on its own platforms rather than losing them to social media. “We’ve seen a lot of transformation in the digital space. We want to get closer to our users, and we want them to visit and return to our platforms,” says James Cooney, vice president of product engineering at Condé Nast.

To support this goal, the company relaunched the Vogue digital app with new community features and live events. But opening the door to user-generated content introduced a new challenge: moderation at scale. Condé Nast needed an intelligent solution capable of handling up to one million images, videos, and text posts while preserving the creative and contextual nuance critical for fashion and culture. “We had manual moderation processes previously, which was expensive, time-consuming, and unable to scale,” says Cooney.

To accelerate its strategic initiative on community engagement, editorial teams were eager to launch Dogue—a photo contest where Vogue app users could upload pictures of their dogs, with the winner featured on a digital Dogue cover alongside celebrity pets. The contest created pressure to accelerate timelines, pushing the engineering team to bring the platform to life faster than expected and put a new moderation system to test.

That’s when Condé Nast turned to AWS. “Condé Nast and AWS have a long and deep relationship, and we always know we can talk to experts and understand the newest technologies,” says Cooney. After consulting with AWS, the company implemented Amazon Nova AI foundation models using Amazon Bedrock.

Amazon Nova Pro, a multimodal model, delivered the accuracy and speed needed to moderate user-generated content at scale while supporting future expansion into video. “It would not have been possible for us to do that without a ton of development work and ongoing operational overhead,” says Cooney. “We can bypass all that with a multimodal approach on Amazon Nova Pro. We just work with the prompt and drop video in, with no pre-processing required.”

Working closely with the AWS account team and generative AI experts, Condé Nast’s engineering teams rapidly iterated on prompts and prototypes. “Our team was able to put together a solution that looked almost production ready after just a few weeks of working with the code and guidance provided by AWS,” says Cooney. Within a month, Condé Nast had a successful proof of concept, followed by just two additional months for full integration, deployment, testing, and refinement.

Condé Nast initially launched Dogue using human moderation while preparing to integrate Amazon Nova Pro. But when submissions quickly exceeded expectations, the company switched to Amazon Nova Pro mid-event—a change that demonstrated both the agility of Condé Nast’s engineering team and the strong support from AWS.

With Amazon Nova Pro, Condé Nast saved two weeks of manual content moderation effort, achieved higher accuracy, and processed submissions faster, delivering the precision and speed needed to safely handle thousands of entries in real time. The system scaled seamlessly to support 56,700 photo submissions, 47,300 comments, and 1,300 replies during the Dogue contest. “Amazon Nova Pro gives us the peace of mind to moderate user submissions at scale with confidence while lowering our risk,” says Cooney.

The project also highlighted the importance of cross-functional collaboration. “On their own these features don’t resonate. But when our editorial teams leaned in and tied it together, the audience response was completely different,” says Cooney. Building on this success, Condé Nast plans to roll out the platform across its brands and explore additional Amazon Bedrock features, such as prompt management, guardrails, and knowledge bases, to create a shared framework for safe and engaging community experiences.