Production Track¶
This track focuses on operational ML: model serving, drift detection, release decisions, and system-level tradeoffs.
Recommended Sequence¶
projects/mlops-drift-production-showcaseprojects/batch-vs-stream-ml-systems-showcaseprojects/model-release-rollout-showcaseprojects/demand-api-observability-showcase
Core Skills Covered¶
- Monitoring feature and prediction drift.
- Serving models with contract-aware API endpoints.
- Canary release decisions and rollback criteria.
- Batch vs stream pipeline tradeoffs.
- Observability with structured logs, metrics, and traces.
Evidence Artifacts To Inspect¶
artifacts/drift/outputs in MLOps showcaseopenapi.jsonin API showcasesartifacts/registry/model_versions.json- rollout decision logs and simulation outputs
Suggested Reflection Prompts¶
- Which KPI would trigger a rollback first and why?
- What drift signal is most actionable for retraining decisions?
- Where should alert thresholds differ between batch and online systems?