Server-side rendering and AI localization: Highlights from our recent Laravel meetup

The event was kicked off by János Horváth, BrokerChooser’s CTO, who has been with the company for six years. Drawing on our experiences, he shared how server-side rendering has helped us solve various issues that were introduced after adapting React.js such as slower page load times and Cumulative Layout Shift. In his presentation, János used informative and interactive examples to show how the team integrated SSR into our technology stack through the open-source package we created. The presentation included a demonstration of a sample code, which helped to show practically how our package compares to default render methods.

The presentation was followed by a panel discussion, moderated by backend developer Zsuzsanna Bedő-Huszár, where Jani was joined by tech leads Imre Nagy and Ádám Sinka. Our tech experts explored a real hot topic: how companies can use machine translation to localize their content. 

The genesis of our translation project was an inspiring previous hackathon experience, where we managed to achieve remarkable progress within two weeks. Earlier encounters with generative AI tools and image generators, as well as experimentation with chatbots, paved the way for our deeper exploration of AI-driven translation. The motivation to localize our content came from our founders, who recognized the limitations of traditional translation methods. By leveraging AI, we sought to automate content updates and eliminate the laborious manual process.

The localization project presented unique challenges. With numerous pages and over 500 data points per broker, manual updates were not a feasible option. Instead, we adopted a variable-based approach, utilizing a process that dynamically reads and updates articles. The translation process covers not only content but also the names of variables, allowing GPT-4 to learn and accurately translate these variables within sentences. We also relied heavily on prompt engineering to help optimize the translation process further. 

The project has already resulted in a significant milestone: we published 10,000 new pages in April and a whopping 26,000 in May. In contrast, before this project, BrokerChooser had only around 5,000-10,000 pages in total. By harnessing the power of AI, our company can now reach consumers in a host of languages, leading to a more personalized and engaging user experience.

After the formal part of the event, participants were invited to stay and mingle with our team members over pizza and drinks, providing an opportunity to chat about the topics discussed in a more informal setting. One participant highlighted the advantage of having an interactive panel discussion after a presentation, as well as the benefits of including hands-on examples. Another guest said he came to motivate himself and was satisfied with the outcome as he received useful advice that he will follow in his future. 

Overall, we were thrilled with the turnout and the outcome of the event, which we consider a huge success. We were able to bring together like-minded individuals to share their experiences and learn from each other, which is a great foundation for further collaboration.