Explore the world of Bayesian machine learning in this 35-minute conference talk from the Data Science Festival Summer School. Delve into the advantages of considering all possible parameter vectors compatible with observed data, providing a robust approach to model uncertainty and allowing for the gentle injection of prior knowledge. Compare this method to classical machine learning techniques, which focus on finding the 'best' vector of model parameters to fit data, potentially leading to overfitting issues. Learn about the benefits and drawbacks of the Bayesian approach, and discover practical applications from Egor Kraev, Head of AI at Wise. Gain insights into handling model uncertainty, avoiding overfitting, and leveraging prior knowledge in your machine learning projects.