DevOps for Data Scientists

Course details

Data scientists create data models that need to run in production environments. Many DevOps practices are relevant to production-oriented data science applications, but these practices are often overlooked in data science training. In addition, data science and machine learning have distinct requirements, such as the need to revise models while in use. This course was designed for data scientists who need to support their models in production, as well as for DevOps professionals who are tasked with supporting data science and machine learning applications. Learn about key data science development practices, including the testing and validation of data science models. This course also covers how to use the Predictive Model Markup Language (PMML), monitor models in production, work with Docker containers, and more.

Learning objectives
Using Git for version control
Incorporating model testing into the deployment process
Working with the Predictive Model Markup Language
Securing the data science models in production
Monitoring models in production
Creating a Dockerfile for data science models

Concepts

0. Introduction

01 - Welcome
02 - Target audience

1. Data Science Development Practices

03 - Data science and software engineering
04 - Collecting and munging data
05 - Experimenting with data, features, and algorithms
06 - Testing and validating models

2. Data Science Models to Production

07 - Version control for data science models
08 - Predictive Model Markup Language
09 - Deploying models with automation tools

3. Deployment Practices

10 - Deploying to staging environment
11 - Canary deployments
12 - Securing the data science models in production
13 - Monitoring models in production

4. Data Science Models in Containers

14 - Introduction to Docker
15 - Creating a Dockerfile for data science models
16 - Data science Docker image repository

Conclusion

17 - Overview of DevOps best practices for data science

DevOps for Data Scientists

Authors

Dan Sullivan