Training 2: Implementing a production MLOps pipeline to deliver end-to-end ML applications at scale (3 hours)
Hands-on Lab
Software Engineer, ML Practitioner, ML Engineer
Ray Data, Train, Tune, ServeIn this course we'll learn how to design, develop, deploy and maintain ML applications at scale. While we will finetune an LLM for a text classification task, our focus will largely be around the model agnostic aspects of production ML. We'll cover the best practices for experimentation, as well as using mature CI/CD pipelines to launch, maintain and iterate (continual learning) on our application.
We will use Ray AI Runtime to implement core ML workloads at scale (data ingest/processing, training, hyperparameter tuning, serving, etc.) and to integrate with mature MLOps tooling for contextual workloads (experiment tracking, data validation, model monitoring, etc.).
By the end of this course, you'll know how to take an ML application from prototype to production and apply it to your own bespoke context, stack and scale.
Learning Outcomes
- Design, develop, deploy and maintain a ML application at scale following the ML and software best practices.
- Learn how to use Ray AI Runtime to implement core ML workloads at scale and to integrate with mature MLOps tooling.
- Extend the experience to your own specific context, applications, stack and scale.
Prerequisites
- Basic familiarity with user-facing functionality of language models like chatbots, answerbots, or translation apps
- Intermediate programming skills with Python
- Basic experience with Ray is helpful but not required
- No knowledge of language model internals is required
