I work for a startup as a data scientist. That means I can work as a data scientist, data engineer, data analyst, or in any other position where the word “data” appears in the title. Although I enjoy my work, I want to cry EVERY TIME I have to work with Sage maker, especially when it comes to generating endpoints.
If you need to follow a well-established technique, the documentation is extensive; nevertheless, if you need to perform a more customized action, things can quickly get out of hand. Right now, I’m attempting to implement a customized vision transformer model that functions flawlessly locally. I get an error as soon as I publish the endpoint, but it doesn’t explain why it’s there. Everything seems like a pretext to charge you for their assistance.
Avoid using Sagemaker at all costs. It accomplishes nothing that open source libraries cannot.
Sagemaker (and related services like Azure ML) charge you more for the identical computing resources that you could obtain on your own. That is all there is to it.
Cloud businesses also attempt to vendor lock you and your team in this way.
I am aware that developing, honing, and implementing your machine learning model is made simpler by learning SageMaker. The most widely used cloud provider globally is Amazon Web Services, and data scientists increasingly need to be as knowledgeable about cloud services as someone in DevOps.
