Course topics

Topic 1: MLops Introduction
  • ML in real world
  • Data Engineering into ML Models
  • Roles & Responsibilities
  • Data Types and ML Architectures
  • Scenarios
  • A different version of MLOps
  • Typical Tooling
Topic 2: End-to-end basic case (Computer Vision)
  • Model in notebook
  • Split training & inference
  • Automatic builds
  • Containerization
  • Training automation
  • Model Generation & Repository
  • Production deployment
  • Integration into existing pipeline
Topic 3: Model preparation
  • Model Versioning
  • Data set visioning
  • Containerization
  • Annotation Tooling
  • AutoML
  • Explainability
  • Feature Engineering
  • Model Repositories
Topic 4: Pipelines & Workflows
  • Release Pipelines
  • AirFlow / Pachyderm
  • Code hooks
  • KubeFlow / KubeFlow Pipelines
  • AWS SageMaker vs Azure ML Studio vs GCP DataLab
  • Feature Stores
  • Re-training & metainformation
  • Data Quality
Topic 5: Production Deployment
  • Containers/Docker/Kubernetes/KubeFlow
  • Inference
  • Monitoring
  • Model Drift
  • Data Drift
  • TF Serve/Seldon/KFServe
  • Server/Edge/Mobile
Topic 6: Processes
  • What can go wrong?
  • Maturity of the models
  • Meta Architecture
  • Scenarios – Operations & ML
  • Google MLOps Maturity Levels

Про факультет

Важлива інформація

Контактна інформація