Build a “For You” recommendation page in an hour with Fastly

“For You” pages have become commonplace within social and web platforms, with TikTok and Instagram featuring them prominently in their product experiences. But, as we explore in our 13th episode of Fastly Developers Live, any platform with an archive of content can use Fastly to build a For You page in just under an hour. 

In the episode, we use the Metropolitan Museum of Art’s Archive API, Fastly’s KV store, and open-source libraries to build a recommendation feed for each unique visitor. We review the metadata across collection pieces, vectorize that data, and use those results to build the foundation of the recommendations. If that seems intimidating, you’re not alone! But don’t worry, we take our time to walk through the process at a high level, and step through to code we use. You’ll come away learning a bit more about machine learning and processing large, complex, disparate datasets.

Finally, to bring the project to life, we use Fastly Compute’s functionality to rewrite HTML bodies to seamlessly stitch the recommendation feed into the Met’s gallery pages in real-time, matching the design aesthetic and user experience. Also, the results are highly cachable, taking advantage of Fastly’s CDN for instant production scale. Check out the full episode below!