Contribute Media
A thank you to everyone who makes this possible: Read More

How you really get your data science models into production the cool way!


PyData SF 2016

This talk discusses one of Pivotal Labs’ core principles, API first, and how this can help to overcome the common language problem between data scientists and software engineers. Topics covered in this talk are:

  • Tools
  • Software engineering methodologies like continuous deployment and TDD
  • Microservices
  • PaaS and Cloud Native Data Science

Over the years we have seen many non-tech companies starting to build their own data science teams because they realize that data will eat the world. In an effort to understand the space, these teams have begun to play with their data and create early prototypes. Unfortunately, those prototypes primarily end up in powerpoint and die. Of the ones that move forward, there is a gap in knowledge of their development team in how to release to production.

From our experience, we’ve seen their data science and development team do not speak a common language. At Pivotal Labs, we found a good way to overcome this language problem. Our solution is to follow an API first approach which is one of our core principles.

In this talk, I want to to share my experiences of how to put these models into production. I will focus on the tools that we use for this and what data science has to do with microservices. My presentation will contain an end-to-end data science example.


Improve this page