Reliable model outputs through intentional prompt design
Module 1
Prompt Design Fundamentals
- How LLMs process prompts: tokenization, context window, attention — the mental model that helps
- System prompts vs user prompts vs assistant prefills
- Zero-shot, one-shot, and few-shot prompting: when examples help
- Chain-of-thought (CoT): getting models to reason step by step
- Output format control: JSON, markdown, structured extraction
- Role prompting: assigning personas and expertise
- Negative constraints: telling the model what not to do
- Prompt length trade-offs: more context vs token cost
- Temperature and sampling parameters: when to adjust and when to leave alone
- Testing prompts: evaluation criteria, expected output, edge case generation
Module 2
Advanced Techniques & Production Patterns
- Tree of Thought (ToT): exploring multiple reasoning paths
- Self-consistency: running multiple completions and voting
- ReAct: reasoning + acting for tool-using prompts
- Prompt chaining: decomposing complex tasks into sequential prompts
- Meta-prompting: using an LLM to generate or improve prompts
- RAG integration: prompt templates that incorporate retrieved context well
- Prompt injection and jailbreaking: understanding and defending against them
- Prompt versioning: treating prompts as code — git tracking, change review
- A/B testing prompts: measuring quality improvements systematically
- Prompt libraries: organizing and reusing prompt templates across the codebase
- Model-specific differences: what works on GPT-4 vs Claude vs Gemini
What your team walks away with
Developers and product teams who can design prompts that produce reliable, high-quality outputs — and build the evaluation infrastructure to measure and improve them systematically.
- Write zero-shot and few-shot prompts that produce consistent structured outputs
- Apply chain-of-thought and self-consistency for complex reasoning tasks
- Version and test prompts systematically like any other piece of code
- Defend against prompt injection and design prompts that degrade gracefully under adversarial input
Book the Prompt Engineering training
A practical one-day workshop — participants build and evaluate real prompts throughout. Works well before or alongside the AI + Agentic Development training.
Get in touch