Data Science, from Concept to Production

by Chris Menezes

machine learning data science devops systems 10 minutes

So you've got an idea for a machine learning product, but how do you actually get it to production? From going on-call for ML models, to ensuring that models built by your data scientists can be used by your engineers, join me for a fast paced guide to the world of data science in production.

Building a great product is hard enough when your system is deterministic. Sprinkle in some probabilities, stochastic data, and unpredictable users? Welcome to the wonderful world of machine learning, where nothing is quite as it seems, and sometimes your models just don't want to behave.

With the proliferation of data science tools, techniques and computational resources, it's becoming easier than ever to experiment with machine learning. However, when you want to go past the experimentation phase and bring a model into a production environment and - gasp - actually have customers interact with it, it's a different ball game.

In this talk, I'll take some time to reflect on and share lessons learned from bringing a machine learning system into production. After coming to this talk, attendees will be better equipped to bring their ML experiments from the lab to their customers."

About the Author

I’m a software engineer focused on data science and machine learning projects. I work at Pagerduty, applying intelligent systems to the world of DevOps with the goal of reducing on-call pain for our customers. Prior to that, I worked as a data visualization consultant on intelligent search systems. I’m a proud graduate of the University of Waterloo’s System Design Engineering program, and I’m working towards my MSc in CS through Georgia Tech over nights and weekends. When I’m not busy with work or school, I’m riding my single speed around Toronto :)

Author website: