MLOps Zoomcamp

Our MLOps Zoomcamp course

• Sign up here: https://airtable.com/shrCb8y6eTbPKwSTL (it's not automated, you will not receive an email immediately after filling in the form)
• Register in DataTalks.Club's Slack
• Join the #course-mlops-zoomcamp channel
• Tweet about the course!

Overview

Objective

Teach practical aspects of productionizing ML services — from collecting requirements to model deployment and monitoring.

Target audience

Data scientists and ML engineers. Also software engineers and data engineers interested in learning about putting ML in production

Pre-requisites

• Python
• Docker
• Being comfortable with command line
• Prior exposure to machine learning (at work or from other courses, e.g. from ML Zoomcamp)
• Prior programming experience (1+ years of professional experience)

Timeline

Course start: 16 of May

Syllabus

There are five modules in the course and one project at the end. Each module is 1-2 lessons and homework. One lesson is 60-90 minutes long.

This is a draft and will change.

Module 0: Introduction

• What is MLOps
• Running example: NY Taxi trips dataset
• Why do we need MLOps
• Course overview
• Environment preparation

Module 1: Processes

• CRISP-DM, CRISP-ML
• ML Canvas
• Data Landscape canvas
• (optional) MLOps Stack Canvas
• Documentation practices in ML projects (Model Cards Toolkit)

Instructors: Larysa Visengeriyeva

2 hours

Module 2: Training

• Tracking experiments
• MLFlow
• Model registry
• ML pipelines, TFX, Kubeflow Pipelines
• Scheduling pipelines (Airflow?)
• Model testing

Instructors: Cristian Martinez, Theofilos Papapanagiotou

Homework:

• ? something with MLFlow perhaps as it’s easier to run locally

Module 3: Serving

• Batch vs online
• For online: web services vs streaming
• Serving models with Kubeflow+Kubernetes (refer to ML Zoomcamp)
• Serving models in Batch mode (AWS Batch, Spark)
• Streaming (Kinesis/SQS + AWS Lambda)

Instructors: Alexey Grigorev

Homework:

• Deploy a model with Spark (local mode)

Module 4: Monitoring

• ML monitoring VS software monitoring
• Data quality monitoring
• Data drift / concept drift
• Batch VS real-time monitoring
• Tools: Evidently
• Tools: Prometheus/Grafana

Instructors: Emeli Dral

Homework:

• ?

Other things:

• Data quality issues
• Alerts

Module 5: Best practices

• Devops
• Virtual environments and Docker
• Python: logging, linting
• Testing: unit, integration, regression
• CI/CD (github actions)
• Infrastructure as code (terraform, cloudformation)
• Cookiecutter
• Makefiles

Instructors: Sejal Vaidya

Homework:

• ?

Project

• End-to-end project with all the things above

Running example

To make it easier to connect different modules together, we’d like to use the same running example throughout the course.

Possible candidates:

• https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page - predict the ride duration or if the driver is going to be tipped or not

Instructors

• Larysa Visengeriyeva
• Cristian Martinez
• Theofilos Papapanagiotou
• Alexey Grigorev
• Emeli Dral
• Sejal Vaidya

Other courses from DataTalks.Club:

• Machine Learning Zoomcamp - free 4-month course about ML Engineering
• Data Engineering Zoomcamp - free 9-week course about Data Engineering

FAQ

I want to start preparing for the course. What can I do?

If you haven't used Flask or Docker

• Check Module 5 form ML Zoomcamp
• The section about Docker from Data Engineering Zoomcamp could also be useful

If you have no previous experience with ML

• Check Module 1 from ML Zoomcamp for an overview
• Module 3 will also be helpful if you want to learn Scikit-Learn (we'll use it in this course)