MLOps

ML Operations (MLOps) describe a best-practice framework for the successful delivery of AI applications. MLOps is modelled on the success of DevOps (development and system operations) found in software engineering and applies them to ML. Our in house MLOps framework provides a flexible application development platform for AI in healthcare.

The application development framework has three major components:

• MLOps server

  Hosts the machine learning (ML) lifecycle management services.

• Project template

  The starting point for any project using this development framework.

• csc-mlops package

  Available on PyPi and is installed by default by the project template, this handles communication between the project and server, performs automated tasks, and includes helper functions and classes to streamline development.

Transparency is a fundamental value when developing healthcare applications; this framework has been developed open-source, exposing it to community scrutiny and collaboration. The diverse nature of applications that can benefit healthcare and the flexibility that requires has been a guiding principle of this project.

The code is available at https://github.com/GSTT-CSC/MLOps.