ML Application Template

Objective

• Build a serverless ML application in AWS that utilises a CI/CD pipeline.
• CircleCI - Test and build.
• Deploying custom Scikit Learn model inside Docker container.

ML Application Lifecycle

Build

• Setup and manage notebook environment using sagemaker and docker.
• Generate and label data.
• Run Terraform.

Train

• Fetch Model Artefact
• Load Training Data

Deploy

• Set up and manage inference clusters.
• Provision public endpoint with API gateway.

Data Driven Architecture

To Do

• Should the api be broken down into two separate endpoints. One for training and one for inference or should they be put together?