Question

Your team trained a churn-prediction model that must score all 40 million active customers once per night and write the results to BigQuery for the marketing team's morning campaigns. There is no requirement for sub-second responses, and the job only runs once daily. You want the most cost-effective and operationally simple serving approach on Vertex AI. What should you do?

Accepted Answer

Run a Vertex AI batch prediction job that reads the customer records from BigQuery and writes scored results back to BigQuery.

Answer

Deploy the model to a Vertex AI online endpoint with autoscaling and have a scheduled job send individual prediction requests for each customer.

Answer

Deploy the model to an online endpoint with a high minimum replica count to guarantee throughput during the nightly run.

Answer

Export the model and host it on a single Compute Engine VM, then loop through customers with a Python script.