Exam AWS Certified Machine Learning - Specialty topic 1 question 268 discussion

Copilot initially said option D, but when you challenge the machine saying that QuickSight is purely a data representation tool and does not analyse anything (as clearly requested in the question) changed its mind, and says B. "Considering the need for both analyzing metadata and generating a report, Option B might be a better fit because SageMaker Studio has the comprehensive tools needed for both tasks: in-depth analysis and reporting."

Option B involves using Amazon SageMaker Feature Store to set feature groups and assign metadata, then using SageMaker Studio to analyze the metadata. While this approach is valid, it may not be the best choice. Studio is for development not for visualisation.

you can input metadata directly in sagemaker feature store, no need for dydb. https://docs.amazonaws.cn/en_us/sagemaker/latest/dg/feature-store-add-metadata.html

Option D is similar to option B but suggests using Amazon QuickSight for metadata analysis instead of SageMaker Studio. While QuickSight is a viable option for visualization, it may require additional configuration and setup compared to using SageMaker Studio, which is already integrated with SageMaker Feature Store.

On second thoughts, D (QuickSight) is also a possible option because feature store parquet files could be queried by Athena and Athena could be used with Quicksight without any development efforts. Would go with option D

B is the correct option because in option D, Quicksight is used which doesn't support parquet files. Sagemaker feature groups created in offline feature store use parquet to store feature values on S3. This question is talking about auditing which makes offline feature store an obvious choice. In order to use Quicksight, there is an additional step to convert feature store parquet file to a supported format (like CSV, JSON, etc.) and hence, it has more efforts compared to creating a dataframe in Data Wrangler and using it for visualizations

The answer should be D, both the sagemaker studio and the quicksight can analyze metadata but Amazon SageMaker Studio is a web-based, integrated development environment (IDE) for machine learning that provides all the tools you need to take your models from data preparation to experimentation to production. and the question is asking about a solution with the least development so it should be Quicksight

A. NO - Amazon SageMaker Feature Store cannot transform B. NO - SageMaker Studio require Python to analyze the metadata C. NO - custom algorithms are dev-intensive D. YES - use built-in functionnalities

Agree D

Por menor esforço de desenvolvimento descartamos Letra A (levantar um serviço gerenciado como DynamoDB normalmente não é a melhor solução), Letra B (não é performática), Letra C (mesmo motivo da A). Logo por eliminação, Letra D está correta.

This solution meets the requirements with the least development effort because it uses Amazon SageMaker Feature Store, which is a fully managed service that makes it easy to store and manage feature metadata. Amazon SageMaker Feature Store also provides built-in functionality for analyzing feature metadata, so there is no need to create custom algorithms or data flows.

https://aws.amazon.com/blogs/machine-learning/controlling-and-auditing-data-exploration-activities-with-amazon-sagemaker-studio-and-aws-lake-formation/ Studio supports Audit

I think it's B