Industrializing data science: a view from the trenches

Over the last few years, many companies have experimented with data science. With potential business value demonstrated by a proof-of-concept, all that remains is industrializing the developed algorithms: building production-grade machine learning applications, connecting them to live data streams and adapting the current business processes. However, in practice, it remains challenging to bring these data science solutions to production, leaving many projects stranded in the proof-of-concept phase. But how do you manage the industrialization of machine-learning systems deliver the promised business value?

In this talk, we will identify the key challenges using real-life examples. We will focus on facilitating collaboration between data scientists and data engineers, integrating this new type of applications with existing IT systems, and establishing involvement of end-user from the business. Finally, we discuss several concrete solutions.