Reimagining wedding planning using Amazon Nova with Düğün.com
Learn how wedding marketplace Düğün.com expanded vendor discovery for couples by using Amazon Bedrock.
Benefits
reduction in p99 search latency after tuning
conversions and click-throughs with video clips
Overview
Planning a wedding requires dozens of high-stakes decisions. To guide couples through these choices, Düğün.com built machine learning (ML) tools, which outgrew its infrastructure. So the company used Amazon Web Services (AWS) to reengineer its data science layer and build new features that enhance the wedding planning experience. The new natural language search and image-to-video features put couples at the center of every decision: They describe the wedding they imagine, and Düğün.com shows them the most relevant venues and gowns. Now, couples get more personalized guidance across a far wider range of vendors—and more of those vendors get seen.
About Düğün.com
Headquartered in Türkiye, Düğün.com is a large wedding marketplace that connects couples with venues, photographers, gown boutiques, and other vendors across 14 countries.
Opportunity | Using AWS to power AI-driven discovery for Düğün.com
About 400,000 couples plan their wedding with Düğün.com each year. For most, it’s a first pass through an expensive, emotional process on a deadline. But search matched the words in a listing rather than what a venue offered, so vendors that fit a couple’s plans went unseen. Couples also began describing their ideal wedding in natural language. Meanwhile, because the company’s ML tools ran on hand-deployed scripts, it rebuilt the same foundation for every project. So Düğün.com decided to modernize its infrastructure on AWS.
Solution | Building AI services that personalize wedding planning
The company rebuilt its data science layer as shared services. “The beauty of AWS is that you share services across projects,” says Ali Saadat, data scientist at Düğün.com. “We built infrastructure once and reuse it on every project. That takes a new project from zero to hero much faster.”
Scheduled pipelines refresh the search catalog weekly and score active vendors for retention and segmentation. The pipelines run on AWS Lambda, which provides serverless compute, and AWS Step Functions, a service for building and orchestrating distributed applications.
Düğün.com built two features on Amazon Bedrock, a service for building generative AI applications and agents. The first is an assisted-search capability that runs on Amazon Nova, a portfolio of AI offerings. Using this feature, a couple describes their ideal wedding in natural language, such as “a garden ceremony for 80 guests,” and the service returns the vendors that fit. The company deployed the capability to a pilot group ahead of a site-wide rollout.
The second is image-to-video conversion, which is powered by Luma Ray 2 and Amazon Nova Reel models. This helps a bride turn a photo into a short clip and see how a gown moves from every angle before booking a fitting.
Outcome | Accelerating search results while surfacing more options
Search results on Düğün.com now reflect what venues offer rather than the words in their listings. The image-to-video feature doubled conversions and click-through rates, and an internal AI assistant reduced repetitive staff questions and improved response accuracy. Structured venue searches return results in 170 ms on average when the system is warm, and the 99th-percentile (p99) latency fell 51 percent after tuning. Using AWS, Düğün.com’s lean team is now free to build what comes next, such as call center summaries and search reranking.
Venue search in natural language
Image-to-video generation for wedding-gown listings
The beauty of AWS is that you share services across projects. We built infrastructure once and reuse it on every project. That takes a new project from zero to hero much faster.
Ali Saadat
Data Scientist, Düğün.comAWS Services Used
More Customer Stories
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages