AWS Certified Machine Learning Engineer - Associate

An ML engineer must visually explore a tabular dataset and apply built-in transformations (imputation, encoding, outlier handling) with minimal code before training in Amazon SageMaker. Which tool is designed for this?

A team wants to store, share, and reuse curated ML features across multiple models and serve them with low latency at inference time, while keeping an offline copy for training. Which AWS capability should they use?

An engineer needs a serverless way to run SQL queries directly against large CSV and Parquet files in Amazon S3 to profile data before training, without loading it into a database. Which service fits best?

A dataset for a binary fraud classifier has 99% non-fraud and 1% fraud examples, and the model rarely predicts fraud. Which data-preparation technique directly addresses this class imbalance?

An ML engineer must build a scalable ETL job to clean and join several large tables and catalog the resulting schema for downstream ML use. Which AWS service is purpose-built for serverless ETL with a data catalog?

A categorical feature 'city' has 3 unordered values (NYC, LA, Chicago). Which encoding is most appropriate so a model does not infer a false ordinal relationship?

Before training, a team must detect potential bias in the training data — for example, whether a facet like age group is imbalanced relative to the target label. Which SageMaker capability computes pre-training bias metrics?

A team needs to create a high-quality labeled dataset for image classification using human annotators and optional automated labeling to reduce cost. Which AWS service provides managed data labeling workflows?