Amazon SageMaker Reviews

We are using Amazon SageMaker to forecast the models. We receive the data into Amazon S3 from the SAP HANA-based systems. Additionally, we are doing preprocessing and sampling for regular data.

The most valuable feature of Amazon SageMaker is its integration. For example, AWS Lambda. Additionally, we can write Python code.

There are other better solutions for large data, such as Databricks.

I have been using Amazon SageMaker for three years.

The stability of Amazon SageMaker is good.

Amazon SageMaker is scalable to the project requirements.

The support from Amazon SageMaker has been positive. We create a ticket with our issues and they contact us with the solution.

I have previously used Databricks. 

The initial setup of Amazon SageMaker is straightforward. The solution is cloud-native making the process take a few minutes. Adding the extensions can take some time. We used CI/CD methods to implement the solution.

We have different teams and we had a team of two DevOps that did the implementation of the solution.

Databricks solution is less costly than Amazon SageMaker.

I rate Amazon SageMaker a seven out of ten.

We use this solution for Outlier Detection using Random Cut Forest. We intend to implement a Predictive modeling project starting in October and have not yet decided on the platform(s) we will utilize.

The challenge for us is balancing the Data Scientists, Technical vs. Analyst.

We are still learning the platform and will conduct more training as we evaluate it for other projects. The few projects we have done have been promising.

The most valuable features of this solution are the Random Cut Forest and the IDE.

I would say the IDE is quite immature, but it is still in its infancy, so I expect it to get better over time.

This is a solution that we have provided to one of our clients.

The client is in the business of consumer goods and they wanted to get accurate demand forecasts to evaluate performance of campaigns and optimize inventory.

The solution is an ensemble of regression models and is deployed on their AWS Cloud and all of the data is on Amazon Redshift. 

The deployment is easy and good. The documentation is pretty good also.

Integration with other AWS services is seamless.

The interface and the IDE could have some improvement. UX isn't bad but could be better.

Orchestration of the ML flow can be made easier (like ETL etc.)

Scalability to handle big data can be improved by making integration with networks such as Hadoop and Apache Spark easier.

Adding certain AI functionalities similar to what DataRobot or Azure AI has would be really great.

I have been using Amazon SageMaker for four to five months.

This is a stable solution. We haven't seen any glitches as of yet.

It is scalable to a degree. We have used several open data sources and have found that for small data, it works well. However, as the volume of data increases, there are issues with respect to scalability.

In general, I would say that for small to medium volumes of data, this solution works well. For bigger data, there is room for improvement. 

We have a team of five people who are using SageMaker.

We did not need to contact technical support because the documentation is good and we have in-house expertise.

I have also used the Microsoft Azure Machine Learning Studio and Databricks, and the interface is a little better with these solutions. The Microsoft solution is really good in terms of user experience.

When it comes to deployment and integrating with cloud services, Amazon SageMaker is better as AWS.

I did not have trouble with the initial setup and I don't think that it was very complex. Overall, I would say that it is good.

We have a few experts here who helped with the implementation. The deployment took about a week to get everything ready.

Two people are suitable for maintenance and support.

My advice to anybody who is considering this solution is to think about using multiple cloud services. This solution is good but for complex business problems and big data, it gets a bit trickier. In terms of deployment, it is a clear winner.

From the cost point of view, it's relatively on the higher side.

Overall, there are a few improvements that I want but SageMaker is pretty good.

I would rate this solution a seven out of ten.