GPT5 Prompt Engineering for structured outputs

"I’ve suspected for a while that the single most commercially valuable application of LLMs is turning unstructured content into structured data. That’s the trick where you feed an LLM an article, or a PDF, or a screenshot and use it to turn that into JSON or CSV or some other structured format."

Simon Willison, prolific LLM blogger and creator of Django, 28 Feb 2025

There are three forces converging in LLMs right now

1 LLMs are becoming much cheaper. For example, OpenAI's o1 pro cost $600 per million output tokens. Now OpenAI's flagship model GPT5 costs $10 per million output tokens, a factor of 60x

2 LLMs are becoming very good at extracting structured data which conforms to a JSON schema, even if it requires a little bit of reasoning ability

3 Open source LLMs are also becoming extremely competitive in terms of output quality

As a result, it has become much easier to extract structured data from text over the last year using LLMs, and there are also a lot more LLMs to choose from for this task.

You can get the most out of this investment by engineering your prompts.

This course teaches you a systematic approach for prompt engineering so as to make the most out of your LLM investment. 

Who should take this course?

Everyone who is interested in extracting structured information using LLMs can benefit from taking this course.

Biomedical NLP

Note: I do NOT have a background in biomedicine, but I have now spent over two years on a volunteer project which requires the use of practical NLP tools to extract information from a popular biomedical dataset and have done so quite successfully.

As a student, you do not need to have a background in biomedicine either to complete this course, but it does help if you have an interest in the subject.