TransferLab:Practical Anomaly Detection

In this section, we will look at the four important perspectives on MLOps: the machine learning (ML) lifecycle, ML responsibilities, ML principles, and the applied AI project management framework. We will find out how these concepts interact with each other and lay the foundation for the following sections.

Here we dive deep into project planning, the first stage of the ML lifecycle. You will learn how to divide an AI project into different phases to minimise uncertainty. You will also learn how to plan AI projects from ideation to implementation by conducting a project planning workshop.

In this section, we will demonstrate various methods for optimising data processing in your organisation by applying best practices during data collection, preparation, and management. Among other things, you will learn about techniques such as ETL/ELT, data versioning, automatic quality checks, data provenance tracking, and data catalogues.

This section focuses on various methods for improving model training and model management. You will understand what an ML experiment is and explore techniques such as experiment tracking, model versioning, and model registries. You will also learn how data scientists and ML engineers can collaborate effectively during the modelling phase.

This section introduces various methods for improving the management, monitoring, and maintenance of AI models. You will learn about different model service patterns and deployment strategies. You will also understand how model performance is affected by changes in data, shifts in covariates, changes in labels, and concept drift.

This section covers the various feedback loops throughout the ML lifecycle and discusses ways to orchestrate execution. You will learn what it means to orchestrate an ML workflow, what types of triggers require retraining of models, and how orchestration tools support collaboration in ML teams.