Sunday 15:45–16:20 in Megatorium

Challenges in building Machine Learning models in production

Elena Sokolova

Audience level:


In this talk Elena will describe challenges that arise when building machine learning models that should run in production to personalize user's search. She will present biases that can be hidden in your observational data and can destroy your model completely and will explain several options on how you can correct for them.

Abstract is the largest travel e-commerce website in the world. To make user's search of the suitable property smooth, data scientists are working hard to predict user's needs by building advanced machine learning algorithms. Based on these models the website's content is adjusted to provide only the most relevant information to the user. Although building classical machine learning models for personalization might seem to be a simple task these days, there are a lot of pitfalls due to enormous number of biases hidden in your data, and incomplete information that you get when you work with observational data. In this talk Elena will describe these challenges in details and will explain how to deal with them in practice.

Subscribe to Receive PyData Updates