Introduction to MLOps
Objectifs
- Understand what MLOps is and how it relates to DevOps
- Know the key aspects of MLOps and the data/model lifecycle
- Familiarize with a few example technologies to build a MLOps pipeline
Prerequisits
- Git knowledge
- Usage of command line
- You must understand DevOps
- Capability to create a CI/CD pipeline on GitLab with different stages
- Basic machine learning knowledge
- Basic data-science knowledge
Evaluation
At the end of this lecture, you will be evaluated based on the following criteria:
- Ability to integrate an ML model into an existing project, following the MLOps principles
- MLOps principles include DevOps principles, like testing, versioning, CI/CD, etc.
- Explain the choices made to create the MLOps pipeline
- Explain the process of new data entering the system to the deployment of a new model
- Answer questions like:
- How can I reproduce a specific model (currently in production or previously trained)?
- How can I compare experiments and choose the best model?
- etc.
Additional resources
This section lists a small selection of resources to get you started with MLOps. You can find an excellent practical guide to MLOps here.
Books
- Introducing MLOps
- Building Machine Learning Pipelines
 |
 |
|---|---|