Large Language Models such as GPT-4o, Claude and Google Gemini provide APIs that can be used to develop features that were almost impossibly difficult to build in the past, spanning areas that include human language understanding, image recognition and structured data extraction.
Building software that uses these APIs reliably and responsibly is a topic with a great deal of depth and a lot of hidden traps.
In this workshop we'll explore a range of proven techniques for building useful software on top of this wildly powerful but unreliable substrate.
Topics we will cover include:
- A review of the best currently available models
- Using multi-modal LLMs to analyze images, audio and video
- Use-cases that LLMs can be effectively applied to
- How to access the most capable models via their various APIs
- Prompt engineering
- Retrieval Augmented Generation (RAG)
- LLM tool usage
- Automated evaluations for LLM applications
- The latest options for running local models.
Participants will obtain hands-on experience of building applications on LLMs. Necessary API keys will be provided.