Study Log: AWS Sagemaker Setup

Pulling from the AI Developer learning path on Zero to Mastery I started the AI Engineering Bootcamp to work with AWS SageMaker. I'm not an expert in AI, Python or machine learning. But this is clearly a skillset that only continues to become more important.

Getting setup with Sagemaker takes a significant amount of time. I used Pulumi to setup the initial AWS resources, the source code for that is here.

Once I got the machine up and running and grabbed a sentiment analysis model from HuggingFace it also setup an S3 bucket and EC2 instance automatically. This is obviously the most uncomfortable part of cloud platforms. The random resources added start to build up over time and it can be easy to forget things. This is why I use Pulumi, but it's still uncomfortable that resources can and will be allocated from within a Jupyter Notebook in Sagemaker.

When I ran pulumi down, the operation failed because there was an app running within the instances I created with Pulumi. I had shut it down, but the files you use are stored with AWS storage (another cost).

Still it will be nice to see how this all comes together as the course promises to setup the model on AWS Gateway and test through postman rather than a Jupyter Notebook. As well as use custom data and models.

Outside of a lot of setup and small charges building up, this feels easy to get going. I think largely the challenge here isn't necessarily in the software development side of things. But rather in the data science side. I'm hoping to learn more about that as I continue.