Amazon SageMaker is a fully-managed service that enables developers and data scientists to quickly and easily build, train, and deploy machine learning models at any scale. Amazon SageMaker removes all the barriers that typically slow down developers who want to use machine learning.
Customer Reviews
Ramavtar M.
Advanced user of Amazon SageMaker1) The flexibility to deploy your own algorithms using Docker which is missing in other big cloud ML platforms such as GCP AI. (This is very useful for its integration to other existing ML tools such as MLflow)
2) Support for A/B testing of ML Models
3) Support for batch processing
4) Support for lambda scheduling
5) A very well managed GitHub sagemaker repository supported by blogs to provide all necessary examples to get started.
6) Support for Reinforcement learning.
7) Support for third party software through AWS market place.
1) The very unclear documentation/support for Models monitoring after deployment over Sagemaker endpoints.
2) Too many APIs to do the same thing. Which might confuse the user in finding which one is the best way.
3) UI can be better.
4) Project Management and adding collaborators feature is missing or not directly supported.
6) The logs tracking through Cloudwatch can be better managed.
7) It should provide an out-of-the-box workflow management solution such as Airflow.
I found Amazon Sagemaker better than GCP AI and IBM Watson for my use case. It was mainly related to having support for custom algorithms. Regarding the existing algorithm such as Linear regression, XGBoost, and Deep Learning all available solutions are more or less comparable. Also, Support for A/B testing of ML Models can be a very useful consideration
Solved problems of Deployment of ML models.
The inbuilt support for Autoscaling and A/B testing proved beneficial for the use cases.
