Workshops: Prompt Engineering

Prompt Engineering for Developers: System Prompts, APIs, Evals

This two-day workshop is for software teams building AI features into their own products. You will learn to design system prompts like code, integrate LLM APIs cleanly with structured outputs and tool use, and measure quality with evals instead of shipping on gut feeling.

Who is this workshop for?

  • Backend and full-stack engineers
  • Tech leads and software architects
  • Data scientists and ML engineers with a product focus
  • Product teams planning LLM features
  • DevOps and platform teams for AI infrastructure

What you will take away

  • You write system prompts with clear structure, versioning and documented assumptions.
  • You integrate the OpenAI, Anthropic, Google and Mistral APIs and know the key parameters.
  • You enforce valid JSON output through structured outputs and schemas.
  • You implement tool use or function calling and handle failure cases robustly.
  • You build an eval set and measure prompt changes automatically.
  • You design for cost, latency, prompt caching and security from the start.

Agenda

  1. Module 1

    Day 1: System prompts as software artefacts

    • Structure: role, context, rules, output format, examples
    • Versioning, reviewing and testing prompts like code
    • XML tags, delimiters and structuring context
  2. Module 2

    Day 1: LLM APIs in practice

    • Message format, temperature, max tokens, streaming
    • Structured outputs and JSON schemas
    • Exercise: an extraction feature with validated output
  3. Module 3

    Day 2: Tool use and retrieval

    • Designing function calling and tool definitions
    • RAG basics: selecting context, citing sources
    • Error handling, retries and fallback models
  4. Module 4

    Day 2: Evals and quality assurance

    • Test datasets, grading criteria, LLM-as-a-judge
    • Integrating regression tests into CI
    • Exercise: an eval pipeline for your own feature
  5. Module 5

    Day 2: Operations and security

    • Cost, latency, prompt caching and model choice
    • Prompt injection, data leakage and logging
    • Data protection in API use and EU AI Act transparency duties

From demo prompt to production-ready feature

An LLM prototype takes an hour. A reliable feature takes far longer: unpredictable input, format errors, model updates and cost all need to be under control. This workshop closes exactly that gap.

Our AI Code Engineers teach patterns proven in real projects and work with your stack and programming language.

APIs and tools

We work across providers and show differences in API design and capabilities. For the development itself we use coding agents.

  • OpenAI API and Anthropic Claude API
  • Google Gemini API and Mistral API (EU)
  • Coding tools such as Claude Code, Cursor, GitHub Copilot and OpenAI Codex
  • Eval and observability approaches for LLM applications

Security and regulation

LLM features open new attack surfaces, above all prompt injection. EU AI Act obligations may also apply, for example transparency for systems that interact with people. We show how to address these requirements technically; for the legal assessment, involve your compliance team.

Frequently asked questions

Which programming language do we use?

Python or TypeScript by default, since both have official SDKs from the major providers. We can adapt the exercises to your stack on request.

Do we need our own API keys?

Ideally yes, so the exercises run in your environment and you can keep the code. Alternatively we agree on a practice environment in advance.

Is this a workshop about coding agents?

No, it is about building LLMs into your own software. For programming with agents such as Claude Code or Cursor we offer separate agentic coding workshops.

Workshop

Get in touch

Shall we build the workshop around your actual AI feature?

Tell us a little about your team – we will get back to you within 24 hours with a proposal for content, format and dates.

Prefer to talk?

We only use your details to handle this request. See our privacy policy.