( reference : Introduction to Machine Learning in Production )
Introduction to MLOps
accurate ML model in Jupyter Notebook
\(\rightarrow\) have to put the model into production,
Will learn skills you need to build & deploy Production ML systems.
Full ML project life-cycle
- 1) scoping :
- 2) data :
- 3) modeling
- 4) deployment
Discipline of building & maintaining production systems,
\(\rightarrow\) “MLOps” ( = Machine Learning Operations )
MLOps Process
Data drift
- distribution of the data you trained \(\neq\) distribution of the data that you’re running inference
key point is “CHANGE”
- world changes & model needs to be aware of that change.
Will deal with…
- building “data pipelines” by gathering, cleaning, validating data sets using TFX + \(\alpha\)
- analytics to address model fairness & explainability issues
- “deployment” : serve the users’ requests