MLOps: Machine Learning in Production

Description

The engineering discipline around models — pipelines, versioning, CI/CD, monitoring, and the operational work that starts after the model is good.

What happens after the model works. It covers reproducible pipelines, versioning data and models together, automated retraining and deployment, monitoring for drift, and the incident practice that keeps ML systems reliable when the world changes underneath them.

Who this is for

• ML engineers whose models keep dying in production
• Platform and DevOps engineers now responsible for ML systems
• Data scientists who need to hand work over without it breaking
• Teams with several models and no consistent way to run them

What you’ll be able to do

• Build reproducible pipelines that run identically for anyone
• Version data, code and models together so any result can be reconstructed
• Automate training, evaluation and deployment with quality gates
• Monitor data drift, prediction drift and business impact
• Roll back a bad model quickly and safely
• Run an ML incident without guesswork

Syllabus — 5 modules, 38 study hours

1. Reproducibility (7 hrs)
2. Versioning data, code and models (7 hrs)
3. Automated training and deployment (9 hrs)
4. Monitoring (8 hrs)
5. Incidents and operations (7 hrs)

Includes a 7-week study plan, practice tasks with stated learning outcomes, an honest readiness self-assessment and a progress tracker.

Independent study material. Completing it does not award a certification or credential.