Presented at PyData NYC on October 18, 2018
Listing on conference website: https://pydata.org/nyc2018/schedule/presentation/42/
The reliability of a data science API for external consumption depends on a good contract, meaning a clearly-communicated definition of a proper request and a proper response. Verifying input and ensuring your response is well-formed is crucial. Learn how jsonschema was used in a balanced team with data scientists and application developers to define the shape and properties of input/output that should be considered valid. This enabled test-driven development and allowed the data scientists and the consumers (i.e. the app developers) to work more independently while not breaking each other’s code. Devs happy, data scientists happy.