Profiling, versioning, and feature stores

Deployment e ciclo di vita in MLOps

Nemanja Radojkovic

Senior Machine Learning Engineer

Data profiling

 

 

Automated data analysis and creation of high-level summaries (a.k.a. data profiles, expectations), used for validating and monitoring data in production

Risk of NOT using data profiles

• Clients complaining, although they submitted erroneous inputs to the model
• No way to identify that data has drifted and our model is no longer valid

 

 

 

 

• Data
• Version
• Control

Let's practice!

Deployment e ciclo di vita in MLOps