How to MLOps: Experiment tracking & deployment πŸ“Š

  • 07-17, 09:30–11:00, Club E
  • 07-17, 11:15–12:45, Club E

All times in Europe/Prague

What's this thing called MLOps? You may have heard about it by now, but never really understood what all the fuzz is about. Let's find out together!

In this tutorial, you will learn about MLOps and take your first steps in a hands-on way. To do so, we will be using Open Source tooling. We will be taking a simple example of Machine Learning use case and will gradually make it more ready for production πŸš€.

We start with a simple time-series model in Python using scikit-learn and first add logging steps to make the performance of the model measurable. Don't worry: we will go through it step-by-step, so you won't be overwhelmed. Then, we will log our ML model and load it back into an inference step. Lastly, we will learn about deploying these actual models by Dockerizing our application πŸ™.

Welcome! You will be learning about MLOps in a hands-on way. So get ready to get your hands dirty and code along! πŸ‘πŸ»

Join us if you 🫡:
- Are working with Machine Learning / Data Science
- Have some experience with Python
- Are interested in MLOps and want to get some hands-on experience
- Are interested in taking your Machine Learning model to production

Contents of the tutorial πŸ“Œ:

  1. [15 min] MLOps: what's the fuzz about?
  2. [15 min] Why Experiment tracking? πŸ“Š
  3. [15 min] Hands-on: Logging metrics with MLFlow
  4. [10 min] Hands-on: Comparing experiments in the MLFlow interface
  5. [15 min] Hands-on: Saving a trained model with MLFlow
  6. [20 min] Hands-on: Loading a model with MLFlow and running inference
  7. [15 min] How to deploy our application? πŸš€
  8. [30 min] Hands-on: Dockerizing our application
  9. [30 min] Hands-on: Deploying our application

🏑 What you will take home

At the end of the tutorial, you will be taking home the following:
- What MLOps is
- When it's applicable, and why it is important
- How you can track your Machine Learning experiments and build better models because of it
- Separate model training from model inference
- Know how you could deploy your ML model to production

❀️ Open Source Software

Many of the used tooling is Open Source. Open software for all!

πŸŽ’ Pre-requisites

Some Python knowledge is required, as well as some general Data Science knowledge: model- training and inference as well as cross-validation. We will not go into details on the Data Science part, but it is good to have a rough understanding about it πŸ‘πŸ».

Jeroen is a Machine Learning Engineer at Xebia Data (formerly GoDataDriven), in The Netherlands. Jeroen has a background in Software Engineering and Data Science and helps companies take their Machine Learning solutions into production.
Besides his usual work, Jeroen has been active in the Open Source community. Jeroen published several PyPi modules, npm modules, and has contributed to several large open source projects (Hydra from Facebook and Emberfire from Google). Jeroen also authored two chrome extensions, which are published on the web store.

Hope to see you at EuroPython πŸ‡¨πŸ‡Ώ! πŸ‘‹πŸ»

Yke is a Machine Learning Engineer at Xebia with a background in astronomy and artificial intelligence. In the industry, he learned that models and algorithms often do not get past the experimentation phase, leading him to specialise in MLOps to bridge the gap between experimentation and production. As a professional in this field, Yke has developed ML platforms and use cases across different cloud providers, and is passionate about sharing his knowledge through tutorials and trainings.

Hope to see you at EuroPython πŸ‡¨πŸ‡Ώ! πŸ‘‹