#tchh22 | Predictive preloads for ultimate web performance?
Talk | IT Engineering & Architecture | 45 min | Englisch
DO | 10:30 | Uwe
The premise is very simple: loading resources that the user needs next before they are needed leads to a big gain in the performance of your web application. So why again isn’t everyone doing that already?
The reason is, that predicting what a user does next is much more complicated and doing it so you can make decisions in the browser during a live user session faces a lot of very interesting engineering and data science questions.
In this talk we look at a predictive preloading system A/B tested with over 100 million users. We explore the machine learning models behind it based on hundreds of millions of user sessions, the achieved performance improvements, and the tradeoffs to consider.
Erik Witt is a Web Performance researcher and VP Product at the research spin off Baqend (https://www.baqend.com/). With a strong background in computer sciences he has led the development of Speed Kit since 2017 to solve web performance for large scale e-commerce players by combining Service Workers with advanced caching algorithms and maschine learning.