MLOps

What you'll be able to do

• Make a training run reproducible months after the fact
• Serve a model behind an API with sane scaling and rollback
• Detect drift before a customer reports it
• Automate retraining without letting it run unsupervised

Modules

1. 01
   The ML lifecycleWhy models rot, reproducibility, environment pinning, the notebook-to-production gap
2. 02
   Data and feature managementData versioning with DVC, feature stores, training and serving skew
3. 03
   Experiment trackingMLflow, run comparison, model registry, promotion between stages
4. 04
   Packaging and servingModel containers, batch versus real-time inference, GPU scheduling, autoscaling
5. 05
   OrchestrationKubeflow and Airflow pipelines, scheduling, retries, dependency management
6. 06
   Monitoring in productionDrift detection, data quality checks, feedback loops, shadow deployment and rollback