The repo is to accompany the Friendly MLOps workshop at ODSC East given by Rajiv Shah and Tim Whittaker.

Rajiv and Tim provide an easy explanation of critical issues around MLOps and get you started deploying a model.

This workshop relies on DRUM, an open source deployment framework, that provides automated testing and flexibility around deployment pipelines. DRUM provides built-in support for a variety of modeling frameworks including Keras, scikit learn, R, H2O, DataRobot, and more. So join us to see how we can making MLOps friendlier.

• Using DRUM for performance testing of models
• Using DRUM for validation of models
• Using DRUM to get a REST API endpoint
• Show ease of swapping models out (different framewokrs - H2O GLM, DataRobot LGMB, Python Catboost, Python XGBoost
• Instrument humility rules

The notebooks included in this repo are best run on Google Colab. Either use the link within the notebook or upload it to github.

This repo contains

• Colab - Friendly MLOps
• data - folder containing data used in the Friendly MLOPs notebook]
• models - folder containing various models trained on 10k diabetes dataset

While this was meant to run in colab, you can use colab_requirements.txt to set up the python environment locally.