0%
Preparing your experience
English 日本語
0 Start Learning

Your Bag

E-learning materials

Advanced Prompt Design & Optimization

Move from ad hoc prompts to engineered ones using decomposition, chain of thought, structured output, and evaluation sets that measure whether a change actually helped.

Material Overview
Advanced Prompt Design & Optimization

This E-Learning Material treats prompting as engineering rather than guesswork. You will learn task decomposition, chain of thought and self consistency, prompt chaining, structured output with JSON schemas, and guardrails against prompt injection and unsafe input. Most importantly you will build a small evaluation set, score competing prompt variants against it, and track accuracy alongside cost and latency, so every improvement is measured rather than assumed.

Curriculum

Learning Path

Each level is enrolled separately. Work through them in order, or start at the level that matches your experience.

  1. Level 01

    Beginner

    Purpose

    Move beyond single questions by planning a prompt before writing it, the way you would plan a brief.

    Outcome

    You can turn a vague request into a structured prompt with clear success conditions.

    What You'll Learn
    • Define the task, the audience, the inputs, and what a good answer looks like
    • Decompose a large request into smaller ordered steps
    • Choose between asking for reasoning and asking only for a final answer

    228 CREDS

  2. Level 02

    Intermediate

    Purpose

    Apply reasoning techniques that improve accuracy on tasks where a first guess is often wrong.

    Outcome

    You can raise answer accuracy on hard tasks without changing the underlying model.

    What You'll Learn
    • Use step by step reasoning and self consistency for calculation and analysis
    • Chain prompts so each stage checks or refines the previous one
    • Add verification passes that catch contradictions before the answer is used

    498 CREDS

  3. Level 03

    Advanced

    Purpose

    Produce machine readable output and defend prompts against untrusted input in real systems.

    Outcome

    You can integrate prompts into software where the output must be parsed and trusted.

    What You'll Learn
    • Specify JSON schemas and validate output before it reaches the next service
    • Handle partial, malformed, and refused responses without breaking the flow
    • Defend against prompt injection when user content enters the prompt

    701 CREDS

  4. Level 04

    Expert

    Purpose

    Optimise prompts against measured results, balancing accuracy, cost, and latency instead of relying on intuition.

    Outcome

    You can prove that one prompt is better than another with evidence rather than opinion.

    What You'll Learn
    • Build an evaluation set with expected answers and scoring rules
    • Compare prompt variants and model choices on the same cases
    • Track cost per task and response time and decide where quality is worth paying for

    1,066 CREDS