Inside Hopper’s AI Infrastructure for travel predictions and recommendations

In this talk, Joost Ouwerkerk, the CTO of Hopper, will discuss our production-level, end-to-end infrastructure and workflows to develop AI for travel predictions and recommendations, which now drive 25% of Hopper’s sales. We’ll explore the continuous data ingest from our users (over 30M downloads), and how it enables AI developers to iterate training new neural network designs across millions of users and validate their behavior over multiple data sets. The obstacles faced by the travel industry, such as limited organic feedback (on average people plan 2-3 trips per year vs. 2-3 movies per night on a platform like Netflix), aren’t limited to the travel ecosystem. We will share how the problems we’ve solved in building this infrastructure for training and inference at scale are applicable to others deploying AI-based services.