PRODUCTION SCENARIO
A Generative AI Engineer must label 4 million product reviews in a Delta table with a sentiment class every night and write the results back to a Unity Catalog table.
Which approach does Databricks recommend for this workload?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Submit the full review table to ai_query in a single SQL statement
Databricks advises submitting the full dataset in a single query rather than manually splitting it into small batches, because AI Functions handle parallelization, retries and scaling. Hand-rolled Python batching and pandas UDF fan-out reimplement that scheduling and lose the built-in retries, and fine-tuning a classifier adds a training and serving lifecycle for a labeling job a hosted foundation model already covers.