Skip to content
Certification Prep

AWS ML Specialty

Data Engineering Domain

8 min+80 XPStep 1/11

Land it in S3 first

Nearly every AWS ML scenario starts the same way: data lands in S3, the data lake, and services like Glue, Athena, and SageMaker read it in place.

Format questions are free points. Columnar formats (Parquet, ORC) make analytics queries faster and cheaper than CSV. Some SageMaker built-in algorithms prefer RecordIO-protobuf, which enables pipe mode — streaming data straight from S3 instead of copying it to the training instance.

insight

Exam shorthand: 'faster queries, lower cost, same data' → convert CSV to Parquet. It comes up constantly.