2 What is MLOps ?

 

This chapter covers

  • Introducing the concept of MLOps
  • Discussion of the challenges associated with ML
  • Distinction and similarities between MLOps and DevOps

Machine learning is often not the end product of an organization, but is the means to an end of selling more books, getting more people to watch new shows or making sure that viewers receive an endless stream of interesting content. ML is therefore a tool that drives business value by improving the customers experience, be it internal or external. At its core then, if a model is unable to deliver business value, it has very little reason to exist or pursue.

This value generation is the primary reason why ML and by extension MLOps is hard. Very few companies truly do research on model development and instead reuse architectures and train/adapt off the shelf models for specific domains and problem sets. Availability of comprehensive open source libraries like Huggingface and OpenMMLab also make modeling trivial. After defining a problem and identifying an architecture to solve the problem statement, the hard questions come into focus. How would the model be trained ? How would data get to the model, how would it interact with the other services, where would it be run and how do we make sure the model is accurate over time?

2.1 ML as a loop

2.1.1 Data collection

2.1.2 Exploratory Data Analysis (EDA)

2.1.3 Modelling and Training

2.1.4 Model Evaluation

2.1.5 Deployment

2.1.6 Monitoring

2.1.7 Maintenance, updates and review

2.2 Why is robust MLOps important ?

2.3 Role of MLOps in a mature organization

2.4 DevOps vs MLOps

2.5 Levels of MLOps maturity

2.5.1 Level 0 - Basic

2.5.2 Level 1 - Intermediate

2.5.3 Level 2 - Advanced

2.6 Summary