Reproducible Machine Learning with Buildkite

MLOps has become a trending topic within the machine learning engineering community. MLOps brings DevOps practices into machine learning, and it’s important that ML models are trained reproducibly in the cloud rather than trained in a script that was run on a single device years ago.

This talk will cover how the team coordinates the training process, including fetching the correct data, running code with the appropriate parameters, serializing the model artifact and propagating the artifact to the appropriate places–making the model available in production.