​​This workshop is for data scientists and machine learning engineers who want to learn how to work with LLMs in production systems. Developers will learn how researchers and engineers work with LLMs, emphasizing connections between concepts like fine-tuning and RAG to the mathematical insights, cloud infrastructure, and APIs used to build and improve these systems. Business-minded executives will deepen their sense of what is possible with LLMs, a relative measure of how much it might cost, what new risks generative AI introduces, and how recent generative AI advances fit into the history of enterprise computing. The primary value of this workshop will be to help teams set reasonable goals and understand how to support your team’s next LLM projects successfully.

​What This Workshop Covers:
Using the Metaflow sandbox, a browser-based VSCode in which we’ve provisioned the full stack of ML infrastructure for you, you’ll learn about:
​- Various proprietary off-the-shelf models and APIs.
​- Open-source LLMs.
- ​Prompt engineering.
​- The Retrieval Augment Generation (RAG) pattern.
​- Vector databases.
​- Building a constantly updating production RAG system with Metaflow.

​This course is not a typical data science workshop in the sense that:
- This is an engineering-focused course, not an ML course. You will get to build actual functional systems, not learn about how to train models or run cross-validation. That said, the course is intended to be accessible to non-engineers: we want to make sure that all (data) scientists can build real-world DS, ML, and AI software systems;
​- Outerbounds provides managed infrastructure for the course, including data storage in S3, Kubernetes as a computing platform, workflow scheduling and event-triggering with Argo, and more - all accessed through Python files and notebooks in a familiar IDE.

All workshop correspondence will occur on Slack: please join the community Slack here and then join the channel #workshop-llm-rag-april-2024