MLOps & Machine Learning End-To-End

Objectives

• Understand the Machine Learning Lifecycle and position MLOps relative to DevOps
• Apply the CALMS framework and craftsmanship principles to Machine Learning
• Scope an end-to-end ML project: ML Canvas, value proposition, risk identification

What's covered

• Machine Learning Lifecycle and positioning MLOps relative to DevOps
• The CALMS framework and craftsmanship principles applied to ML
• ML Canvas, value proposition and risk identification

Objectives

• Apply TDD to ML: unit tests, pipeline tests, versioning and packaging
• Master deployment strategies: batch, real-time, edge
• Understand the Feature Store and serving at scale

What's covered

• TDD applied to ML: unit tests, pipeline tests, versioning and packaging
• Deployment strategies: batch, real-time, edge
• Feature Store and model serving at scale

Objectives

• Monitor data (data drift, concept drift), infrastructure and model performance
• Set up feedback loops and automatic retraining criteria
• Orchestrate ML pipelines and manage artifacts (Model Registry, versioning)

What's covered

• Monitoring: data drift, concept drift, model performance and infrastructure
• Feedback loops and automatic retraining criteria
• ML pipeline orchestration and artifact management (Model Registry, versioning)

Objectives

• Implement data versioning (DVC, LakeFS) and experiment tracking (MLflow, W&B)
• Build an ML-specific CI/CD with an adapted testing pyramid
• Capitalize on past experiments via a knowledge repository

What's covered

• Data versioning (DVC, LakeFS) and experiment tracking (MLflow, W&B)
• ML-specific CI/CD and adapted testing pyramid
• Knowledge repository: capitalizing on past experiments