Data science once involved working with a large data set in relative isolation and producing a static report to present at a quarterly meeting. Decision makers in companies have been inspired by the types of insights the data can offer and are asking more questions of it, such as what happens if I change this variable or where did this result come from. Data scientist regularly now present their analysis not in reports but in web apps or behind APIs that allow this kind of interrogation. We have been building data science solutions for 10 years and have learnt many lessons along the way. In this talk we will run through how to build performant, scalable and validated data science web apps, using python.