Getting started: MLOps engineers

Deployed models in production. What the role can do and a first-week path through the screens you will use.

For: MLOps Engineer · 1 min read

Go to the screen:Model operations Platform health Workflow studio Retrain in the sandbox Support tickets

Your role on the platform

You run the models that are in production: endpoints, monitoring, releases and rollbacks, all in Google Cloud Malaysia. When drift crosses a threshold set at validation, the drift response workflow opens a case and you carry out the technical side.

What you can do

Vertex AI pipelines, endpoints and serving configuration
Monitoring, drift thresholds and alerting
Retraining triggers and scheduled runs
Versions, promotion and rollback

Your first week

1Day 1: open Model operations and learn the registry: stage, endpoint, latency, error rate and drift posture per model.
2Day 2: read the alerting rules and check who each alert routes to.
3Day 3: walk through the drift response workflow in the Workflow studio, including the checkpoint where the business owner accepts the retrain cost.
4Day 4: rehearse a retrain in the sandbox on ETB Conversion Propensity and compare the candidate with the champion.
5Day 5: check Platform health and the support tickets routed to you (Platform / technical).

Screens you will use most

Model operations Platform health Workflow studio Retrain in the sandbox Support tickets

Who to ask

Platform Engineering on-call for incidents; the model owner and Model Risk for anything that changes a validated model.