To improve user retention and product satisfaction by providing more relevant gif contents to user search queries while simultaneously reducing the cost and time-to-market for product components such as facial recognition and sentiment analysis.
- Complete backend re-architecture using high velocity microservices in Scala. Refactor the existing monolithic PHP application that rapidly outgrew its original architectural capabilities.
- Built Tag Prediction and Celebrity Detection API.
- Integrated a Moments in Time model, recognizing actions on the gif as an additional search signal.
- iOS and Android Giphy SDKs powering smartphone’s keyboard, all social apps, and native performance on mobile.
- Cutting-edge ML for analytics.
- Reimagined top 3 search engines experiences by using deep learning and AI.
- System capable of handling 100B requests/mo. Also: Improved performance of blacklisting and whitelisting algorithm that sped up overall Giphy API performance by 18%.
- Reduced number of timeouts for Giphy API by 90-95%.
- Improved search service speed by 15%.
A picture’s worth a thousand words — but what about a moving picture? The rise of social media brought about the unintended revival of graphics interchange format (GIFs), and it has become a popular form of self-expression and communication among Gen Zs and millennials. It is estimated that 71 percent of all mobile users use GIFs in their daily communications.
Giphy is a leading online database and search engine that helps people express themselves and make their everyday conversations more entertaining. Giphy capitalized on the opportunity — and arguably pioneered the revival of the image format — by allowing users to search and share desired GIFs instantaneously, giving birth to a new form of communication.
Discover how Proxet’s expertise in machine learning (ML) contributed to Giphy’s search engine’s transformation.
Challenges in Understanding Search Intent
Giphy started out as an image repository for GIFs that allows users to look up GIFs through basic queries. However, as the demand grew, challenges arose. Their PHP architecture had difficulty handling high-load requests and returning results that matched user intents. “The strategy was to throw stuff at the wall and see what stuck, but it was a lot of forward pressure on the engineering team,” said Randy Shepherd, CTO at Giphy.
In the ever-competitive tech world, being able to understand user intent is often the key to market dominance. Failure to do so translates into poor user retention and product satisfaction. Higher demand, with millions of requests at the same time, also posed significant problems for Giphy’s legacy system and a solution was needed to address these issues.
Complete Back-End Re-Engineering to Enable Growth
Proxet partnered with Giphy to re-engineer the infrastructure and migrate the system to high-velocity microservices in Scala to accommodate increasing demand and transform the search engine system. With the expertise of Proxet’s ML engineers, Giphy developed a recommendation system to increase the discoverability of users’ content; Through the use of artificial intelligence (AI) and ML, Proxet’s engineers also developed taxonomy, which allowed Giphy to categorize search queries. Being able to understand which area a particular query belongs to allows the team to group content and search into clusters in order to perform data analysis in structured groups.
“From a deliverable standpoint, we built an ML/AI team that’s doing some really cutting-edge work,” said Anthony Johnson, former CTO at Giphy.
With the new Scala microservices, Giphy’s system is now capable of handling 100 billion requests a month with an increased search service speed of 15 percent. “At the time it was difficult to find someone who had done high-throughput systems,” said Johnson. “Back then they really tended to come from ad tech, and Proxet probably had eight years of experience at that point in high-speed systems through ad tech experience that was more or less directly transferable.”
“You have to balance out the progression of the platform [...] so that you don’t fall apart,” added Shepherd. “Proxet has largely enabled that because we staff folks from Proxet, not only in product teams but also in our infrastructure.”
Growing Market Dominance
Proxet’s engineers also helped migrate Giphy’s third-party integrations, sustaining Giphy’s high-load usage and contributing to its widespread adoption. By adopting a high-performance Remote Procedure Call (RPC) framework (gRPC), Giphy is able to keep its services fast, stable, and fault-tolerant as they scale to over 10 billion pieces of content a day. Along with Giphy’s native engineering teams, Proxet engineers helped provide integration tools for platforms such as Facebook, Instagram, Slack, and Twitch, as well as extensions for different browsers with a renewed microservice approach, greatly facilitating performance and scalability.
With the expertise from Proxet, Giphy now enjoys a refined infrastructure that can accommodate faster deployments and monitoring, as well as improved functionalities that translate directly into better user experience and retention.
Bringing On Skill Sets That Were Hard to Get
“We were able to go from a New-York-only company to having different time zones, expanding our ability to hire, bringing on skill sets that were hard to get, especially in the New York area, and do so much more rapidly than we would have ever been able to do in New York.”
Proxet’s partnership with Giphy goes well beyond team augmentation. Proxet’s engineers are well-integrated into Giphy’s engineering teams and share their knowledge and expertise with their teams. “When you expand a team that’s very close physically and socially, it’s very challenging,” said Johnson. “But I think over the years the team has become integral to what Giphy ships.”
“We were able to go from a New-York-only company to having different time zones, expanding our ability to hire, bringing on skill sets that were hard to get, especially in the New York area, and do so much more rapidly than we would have ever been able to do in New York,” Johnson added. “Small groups of incredibly skilled people like with Proxet are very difficult to find otherwise.”
Contact Proxet today and discover how Proxet’s expertise can expand your capabilities.