Reducing Search Time by 75% Using Amazon Bedrock with Bynder

Bynder initially acquired a company to incorporate visual search capabilities into its software. The fast rate of customer adoption encouraged Bynder to build on its success. At the 2023 AWS re:Invent conference, Bynder realized that the newly released Amazon Titan Multimodal Embeddings would empower its team to continue to innovate and deliver AI capabilities at an even faster rate.

Using the Titan Multimodal Embeddings foundation model in Amazon Bedrock, businesses can power more accurate and contextually relevant multimodal searches. During the keynote presentation, Roald Bankras, director of system architecture at Bynder, began testing the API—and quickly got it up and running. “Amazon Bedrock with Titan Multimodal Embeddings solved the puzzle for us,” says Bankras. “We could very easily turn this into a scalable solution.”

Using Amazon Bedrock, Bynder extended its comprehensive AI-powered search capabilities in its software. The solution converts images into vectors using Titan Multimodal Embeddings so that customers can find assets by selecting similar images or describing what they’re looking for in natural language.

When users input a query, the platform processes it through Amazon Bedrock. Then, both the search terms and the stored images are converted into vectors to identify matches according to visual and contextual similarity. “The setup using Amazon Bedrock made it very easy to get those vectors because it was a very clear API,” says Bankras.

One of Bynder’s customers states that the time spent searching for assets for a typical campaign task has decreased by 75%, resulting in a big boost in productivity for the team. The search results are also deeper and more accurate, returning approximately 50 percent more options on average. Most importantly, the solution scales effortlessly across customers’ massive asset libraries, with virtually no limitations on image quantity.