Module 2 · Prompt Engineering

OBJECTIVE: learn how to talk to an LLM so it gives correct, consistent and useful answers — the most used skill in generative AI. WHAT WE COVER: • Chat messages: system, user and assistant roles • Settings: temperature and max tokens • A clear prompt structure: role, task, context, rules, output format • Zero-shot and few-shot prompting (giving examples) • Asking the model to think step by step • Prompt chaining (splitting a task into steps) and self-consistency • Prompt templates you can reuse in code AFTER THIS MODULE YOU CAN: • Write prompts that give reliable answers in the format you need • Improve a weak prompt step by step and show the improvement • Keep prompts as reusable templates in your code MINI EXERCISE: improve one prompt in 4 versions and measure how many of 20 test inputs each version gets right.