The development phase

MLOps for Business

Arne Jonas Warnke

Head of Emerging Curriculum

Important steps in the development phase

Development phase

Data preparation or feature engineering

Data preparation

Data preparation or feature engineering

Here,

Make those available in a central database
Format data consumable by ML models
Feature Engineering
- Grouping data
- Replace missing data
- Deal with extreme observations

People involved

Data engineer, data scientist, and business expert

$$

A sheet with numbers

Model training or experimentation

Model development and training

Model training or experimentation

Tasks:

Train and tune machine learning model
Compare against alternatives
Assess their performance

Important

Automatically log all results

Task of

Data scientist / machine learning engineer

$$

A hybrid of a human and robot writing code

Model evaluation

Model evaluation

Model evaluation

$$

A human evaluating code

We check here

Are the business requirements met?
- e.g., data privacy
Model stress test
- simulate extreme conditions
Model behavior
- fairness?

$$

Task of

Data scientist / machine learning engineer

Testing and verification

Testing and verification

Testing and verification

Apply software engineering

Tests and
Best practices

to the ML model, e.g.,

Model has no harmful impact on broader system

Task of

Software engineer / machine learning engineer

Testing code

Let's practice model development

MLOps for Business

Preparing Video For Download...