It Worked in the Notebook: Now What?

Moving machine learning from “it worked in a notebook” to something reproducible in production is where things get interesting… and occasionally humbling. In this session, we’ll share our early experience using AWS SageMaker for production ML model versioning and AWS-hosted MLflow for experiment tracking, with a special look at what we learned while migrating parts of our workflow from scikit-learn to PyTorch.

Aimed at teams and practitioners getting started with production ML on AWS, this talk focuses on practical lessons learned: what helped, what added complexity, and what we would do differently next time.