Delivering the ultimate Tinder Swipe Night experience by leveraging personalization at scale
Authored by: Danielle Zegelstein, Engineering Manager | Serge Vartanov, Staff Software Engineer | Jeff Glasse (The Belated Engineer), Senior Engineering Manager | Cooper Jackson (cajaks2), Senior Site Reliability Engineer
At Tinder, our drive to innovate stems from the love we have for our members. Our greatest aspiration is to create experiences that not only surprise our members but delight them as well. So when Swipe Night came around again, our team put our minds together to design the ultimate experience.
Go big or go home
We wanted to make this year’s Swipe Night bigger and better to create a really unique Tinder experience. What if the member could actually be part of the Swipe Night experience? The idea was to embed the members’ Tinder profiles directly into the experience so that they could have an authentic experience. Personalization at scale meant that we had to come up with a solution (nicknamed “the Inception Service”) that could reasonably render millions of videos containing Tinder profiles simultaneously without a hitch.
With 20M members tuning in last year, we did not think it was feasible, but another team member, who had deep knowledge in video post production, was so intrigued by the idea, he went back to his FX skills to build out a quick prototype. With a month until launch, this was an all-hands-on-deck project. Our first challenge was finding a solution for building at scale.
A familiar solution reimagined
Out of the gate, we were presented with three key challenges:
- Building within budget and delivering at a reasonable cost
- Rendering and transcoding a personalized video within the 10 minutes it takes to get through Episode 1
- Ensuring video production is never duplicated so that we don’t accidentally create the same video twice.
We expected millions of Tinder members to log in to this event. Delivering a personalized video within 10 minutes for every member would be computationally expensive. Building a scalable solution that was economically viable was going to be a challenge. While the prototype had been built using Adobe After Effects, we needed open-source software to help us achieve our goal. After a bit of research, we arrived at Blender, an open-source 3D computer graphics software toolset. Historically, the software was used by designers and visual effects artists. This would be the first time we used Blender at Tinder as a backend micro-service running in the cloud.
For our use case, we were able to containerize Blender and put it into the cloud and then get thousands of pods running in parallel. Then we would feed personalized videos from Blender into media convert, which allowed us to transcode them. This enabled us to create personalization at scale in near real-time.
Engineering to the MAX as a team
Most teams are well familiar with the concept of an MVP or “Minimum Viable Product.” But, on the Z Team, when we set out to create something, we like to define tiers of features. These tiers are comprised of Minimum Viable Product (MVP), Minimum Lovable Product (MLP), and Maximum Achievable eXperience (MAX), which contains all of our stretch goals.
The “Inception Service” was completely detached from anything that Tinder had built in the past, but it allowed us to lean in on our strengths around scalability and collaboration. The first prototype was workshopped collaboratively across the engineering organization working remotely. From there, we started regular meetings to bug bash and consult with cross-discipline engineers to tackle this technical challenge we hadn’t seen done before in our industry.
The result was an architecture that surpassed our initial vision. Open engineering allows us to take the best ideas and solutions to build a MAX product ultimately. Further, our partnership with teams outside of engineering opened us up to ideas and concepts that heightened the member experience. Without the collaboration of our cross-functional teams, we could not continually push for innovation. What makes the team unique is that, even on aggressive deadlines, we rarely settle for either MVP or MLP. If it is at all humanly possible, we will find a way to deliver the MAX experience for our members.
Going the distance
It’s always fun to reflect on projects that stretch the engineering skills of the team. We were really excited to be able to make this kind of impact and space for innovation at a company with the scale of Tinder. One of the biggest takeaways was the camaraderie of our teams at Tinder at large. This wasn’t just an engineering project or a Z Team project; everyone stepped in and contributed. We had so many people roll up their sleeves to help us make this project possible.
Are you motivated by creating cutting-edge and novel tech solutions? Check us out; we’re hiring!