Deploying Scalable Machine Learning for Data Science

Learn how to use design patterns for scalable architecture and tools such as services and containers to deploy machine learning at scale.

Introduction

• Scaling ML models
• What you should know

1. The Need to Scale ML Models

• Building and running ML models for data scientists
• Building and deploying ML models for production use
• Definition of scaling ML for production
• Overview of tools and techniques for scalable ML

2. Design Patterns for Scalable ML Applications

• Horizontal vs. vertical scaling
• Running models as services
• APIs for ML model services
• Load balancing and clusters of servers
• Scaling horizontally with containers

3. Deploying ML Models as Services

• Services encapsulate ML models
• Using Plumber to create APIs for R programs
• Using Flask to create APIs for Python programs
• Best practices for API design for ML services

4. Running ML Services in Containers

• Containers bundle ML model components
• Introduction to Docker
• Building Docker images with Dockerfiles
• Example Docker build process
• Using Docker registries to manage images

5. Scaling ML Services with Kubernetes

• Running services in clusters
• Introduction to Kubernetes
• Creating a Kubernetes cluster
• Deploying containers in a Kubernetes cluster
• Scaling up a Kubernetes cluster
• Autoscaling a Kubernetes cluster

6. ML Services in Production

• Monitoring service performance
• Service performance data
• Docker container monitoring
• Kubernetes monitoring

Conclusion

• Best practices for scaling ML
• Next steps