Not long ago, the biggest focus in AI/Machine Learning (ML) was research & development. Since ML has become mainstream in many digital applications, there has been a shift in focus to productionisation of ML models and products.
Business stakeholders invest in ML productionisation as a catalyst for producing real value from ML work. For instance, having a ready to use API to extend ML results into production can be a strategy to realise value and gain traction.
ML CI/CD is one of the best practices to deliver successful ML productionisation. What is it, and why is it important?
It is a set of practices, toolings, systems and platforms to support iterative and continuous development and delivery of ML.
Benefits of establishing CI/CD practice in ML teams include an easier end-to-end path to production, improved quality and productivity, and de-risking key issues earlier.
Examples of toolings in the ML CI/CD setup
Similar to CI/CD in software engineering, toolings play a key part to be built in the CI/CD process. One of the key tenets of ML CI/CD is to enforce good governance and guardrails through automated toolings. As illustrated in the diagram above, the benefits are more than just automation, but also include better model monitoring and observability, stronger security, traceability and PII management, better code hygiene, more consistent team approach, which should lead to better development velocity.
Disclaimer: The statements and opinions expressed in this article are those of the author(s) and do not necessarily reflect the positions of Thoughtworks.