Great models are merely a part of the machine learning revolution

From object detection in images to beating humans in Go and Starcraft, machine learning based systems can solve problems previously unsolvable by classical algorithms. In order to make a true impact on human lives, these systems need to be exposed to end-users in a usable form. I will walk you through the path from a potent machine learning system to a product used by scientists in their daily work. We will go through the process of building such product and the challenges that lie on the way. I will argue that with the correct approach and team, it is possible to deliver complex, specialized ML-based solutions as simple, easy-to-use software. These features help the subject-matter experts to solve real-world problems, rather than spend months learning about machine learning.