Automating DevOps for machine learning

You’ve spent hundreds of hours cleaning your data, engineering features, and training and tuning your model to pinpoint accuracy. But now it’s time to deploy your model into production.

Whether you lead a team of data scientists or are one yourself, you know how time-consuming and problematic deployment can be. Language and environment incompatibilities, manual and duplicative processes, out-of-control costs, and poor communication can destroy all the work you’ve put into building models and slow your machine learning efforts.

Learning common deployment architectures for machine learning in the real world and understanding how to build your model for a production environment can help you avoid pitfalls when scaling up. Diego Oppenheimer (Algorithmia) discusses common problems and solutions and shares best practices from leading organizations that have solved the deployment headache.

This session was recorded at the 2019 O'Reilly Strata Data Conference in San Francisco.