Agenda — Prompt Engineering¶

Session length: ~3 hours | Difficulty: Beginner → Intermediate | Coding time: ~1.5 hours

Why this session matters¶

Prompt engineering is the highest-leverage skill in LLM application development. A well-crafted prompt turns a mediocre model into an excellent one — and a poorly written prompt turns an excellent model into a mediocre tool. Before reaching for fine-tuning or RAG, exhaust what good prompting can do.

Learning objectives¶

By the end of Part 2 you will be able to:

Write zero-shot and few-shot prompts that produce consistent, structured output
Apply chain-of-thought reasoning to improve accuracy on multi-step tasks
Design system prompts that shape model persona, scope, and output format
Extract structured data using both prompt-based and API-native methods
Recognize and fix the 10 most common prompt failure modes

Session outline¶

Time	Topic	File
0:00 – 0:35	Zero-shot and few-shot prompting	01-zero-shot-and-few-shot
0:35 – 1:10	Chain-of-thought prompting	02-chain-of-thought
1:10 – 1:40	Structured output — JSON and schemas	03-structured-output
1:40 – 2:10	System prompts — design and gotchas	04-system-prompts
2:10 – 2:30	Advanced prompt patterns	05-prompt-patterns
2:30 – 2:45	Practice exercises	06-practice-exercises
2:45 – 3:00	Interview questions review	07-interview-questions

Prerequisites¶

Completed Day 1 Part 1 (How LLMs Work)
OpenAI and/or Anthropic API key

Setup¶

pip install openai anthropic pydantic

import os
import openai
import anthropic

openai_client = openai.OpenAI()        # reads OPENAI_API_KEY
anthropic_client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY

The prompt engineering mindset

Think of the model as an extremely capable intern on their first day. They're smart, they want to help, they'll do exactly what you say — including the parts you didn't intend. Be explicit, be concrete, and always specify what you don't want as well as what you do.

Week-01/Day-01-Part-1-How-LLMs-Work/06-interview-questions | Next: 01-zero-shot-and-few-shot