MLOps

Description

AI offers solutions to problems that were previously complex. We can deploy models to perform a task more simply or accurately, automatically detect anomalies in massive datasets, or support individuals with audio/visual impairments in their communication.

For all these goals, a robust process is required. We need to be able to trust that our models are of good quality and can distinguish between models. Quickly and stably updating a model based on new information is essential. To truly benefit from AI, we need to treat our ML solutions just like our software.

This hands-on training teaches you how to develop a model with repeatable processes. You'll learn to apply proven DevOps techniques within a machine learning solution. You register all aspects of your training and experiments in a structured way. You version your model so that a new version is only taken into production if it performs well. You learn about orchestration, such that tasks are executed in the correct order and at the right time. You learn how to deploy, test, and monitor a model in production. All relevant components are stored with version control, making everything reproducible and shareable. You apply everything you've learned by building a proof-of-concept machine learning solution yourself. The training concludes with a case study in which you bring together the learned material.

Prior Knowledge

Data Scientist/Engineer: Experience with ML. We can teach them CI/CD through MLOps. Developer: Have experience with CI/CD. We can introduce MLOps as an add-on.