In this issue:

• we discuss Model Tracing and Lineage;  

• we explore MLTrace, a reference architecture for observability in ML pipelines; 

• we overview M3, a platform that powers time-series at Uber.     

💡 ML Concept of the Day: Model Tracing and Lineage  

Recent issues of our series about MLOps have touched on areas such as monitoring (Edge#141) and observability (Edge#145). Today, we would like to cover a related but differentiated aspect of MLOps pipelines. We are referring to model tracing and lineage management. The relevance of tracing and lineage in ML models is directly related to the unique nature of the lifecycle of ML models. In a typical ML solution, teams will run hundreds of experiments with many different hyperparameter configurations and datasets before settling on a specific nature of a model. After a while, it becomes incredibly difficult to trace back the model's behavior to a particular experiment or data asset.  ...