SageMaker Sandbox

AWS SageMaker has a fantastic set of functional components that can be used in concert to setup production level data processing and machine learning functionality.

• Training Data: Organized S3 buckets for training data
• Feature Store: Store/organize 'curated/known' feature sets
• Model Registery: Models with known performance stats/Model Scoreboards
• Model Endpoints: Easy to use HTTP(S) endpoints for single or batch predictions

Why SageMaker Sandbox?

• SageMaker is awesome but fairly complex
• Spider lets us setup SageMaker Pipelines in a few lines of code
• Pipeline Graphs: Visibility/Transparency into a Pipeline
  • What S3 data sources are getting pulled?
  • What Features Store(s) is the Model Using?
  • What's the Provenance of a Model in Model Registry?
  • What SageMaker Endpoints are associated with this model?

Installation

pip install sagesand