Deployment and orchestration of ML workflows
Practise only this domain. 36 questions with explanations, 20 at a time.
Start this domainSample questions
- A team wants a repeatable ML workflow that preprocesses data, trains a model, evaluates it, and registers the model only if its metrics pass a threshold. The whole flow should be defined as a directed graph and integrate natively with the model catalog. Which service is best?Medium
- A team wants to send a small first slice of production traffic to a retrained recommendation model, watch error-rate and latency alarms during a bake period, and then shift all remaining traffic at once if everything looks fine. Which SageMaker traffic shifting mode is described?Medium
- An inference system must run a feature-normalization container followed by a classifier container, so each request passes through both in order behind one endpoint. Which SageMaker deployment option matches this?Hard
- A trading platform needs model inference to survive the loss of an entire AWS Region, with automatic failover and no manual runbook. Which architecture should be used?Hard