Conf42: Machine Learning 2021

Developing high-quality machine learning models involve many steps. We typically start with exploring and preparing our data. We experiment with different algorithms and parameters. We spend time training and tuning our model until the model meets our quality metrics, and is ready to be deployed into production. Orchestrating and automating workflows across each step of this model development process can take months of coding.

In this session, I show you how to create, automate, and manage machine learning workflows using Amazon SageMaker Pipelines. We will create a reusable NLP model training pipeline to prepare data, store the features in a feature store, fine-tune a BERT model, and deploy the model into production if it passes our defined quality metrics.