Skip to content
Neural Soft
Contact us Español (View this page in Spanish)

Building with AI

Prompt engineering course for technical teams

Prompts that behave the same in production as in the demo.

Request information

A prompt that works in the chat does not always work in production: change one input and the answer comes back in a different format, the model ignores an instruction or the cost shoots up. This course treats prompt engineering as engineering: with structure, testing and version control.

We work on the techniques with the most impact in real applications: robust system instructions, structured outputs that can be validated, well-chosen examples, reasoning, long-context management and caching. And we finish with what is usually missing: sets of test cases so you can change a prompt without fear.

What you will learn

  • Write clear, robust and maintainable system instructions
  • Get reliable structured outputs to integrate into code
  • Use examples, reasoning and task decomposition
  • Manage context and cost in real applications
  • Version and test prompts like any other code

Syllabus

8 modules, from the fundamentals to the final project.

  1. How a model processes a prompt

    • Tokens, context window and temperature
    • Differences between models and providers

  2. Anatomy of a good prompt

    • Role, context, task, constraints and format
    • System instructions versus user messages
    • Delimiting data with tags

  3. Examples and reasoning

    • Few-shot prompting and choosing examples
    • Asking the model to reason before answering, and reasoning models
    • Breaking tasks into prompt chains

  4. Structured outputs

    • JSON with a schema and validation
    • Reliable extraction and classification
    • What to do when the output is invalid

  5. Long context and cost

    • Ordering long documents within the prompt
    • Prompt caching
    • Choosing the right model for each task

  6. Prompts for code and QA

    • Generating, reviewing and explaining code
    • Test cases and synthetic data

  7. Test and version

    • Sets of test cases
    • Comparing versions of a prompt
    • Prompts in the repository and in continuous integration

  8. Security

    • Instruction injection
    • Filtering input and output

Tools you will practise with

  • OpenAI, Anthropic or Google APIs
  • Python or TypeScript
  • promptfoo

Hands-on project

You design, test and version the prompt for a real task in your team (extraction, classification or generation), with a set of test cases that measures its quality before and after each change.

Requirements

  • Basic programming in any language.

Other courses

Related articles

Want this course for your team?

We adapt the content, length and format to your company.

Request information