Day 7: Prompt Engineering + Week 1 Checkpoint
Learning Objectives
- - Use Bedrock Prompt Management for centralized prompt templates
- - Build Prompt Flows with conditional branching
- - Implement prompt versioning and governance with CloudTrail
- - Apply advanced prompting patterns (chain-of-thought, JSON Schema enforcement)
- - Validate learning with the AWS Official 20-question Practice Set
Tasks
Tasks
0/6 completed- Read30m
Amazon Bedrock Prompt Management Documentation
Centralized prompt templates, versioning, parameterization. Eliminates prompts in code.
- Read30m
Amazon Bedrock Prompt Flows Documentation
No-code workflow builder for sequential prompt chains with conditional branching.
- Read30m
Prompt Engineering Guidelines for Bedrock
Best practices for prompting each model family on Bedrock.
- Blog25m
Implementing Advanced Prompt Engineering with Bedrock
Chain-of-thought, few-shot, structured output patterns.
- Hands-on60m
Build a Bedrock Prompt Flow with Conditional Branching
Hands-on notebook for Prompt Management and Flows.
- Practice Exam60m
AWS Official 20-Question Practice Set (Week 1 Checkpoint)
Score yourself. For every wrong answer, note which skill number it maps to.
Exam Skills
Write your understanding, then reveal the reference answer.
Hands-On Lab
Build real muscle memory with these activities.
Create a Prompt Flow with Conditional Branching
Build a Bedrock Prompt Flow that routes different query types to specialized prompts.
- 1 Open Bedrock console → Prompt Management → Create prompt
- 2 Create two prompt templates: one for 'summarization' and one for 'question-answering'
- 3 Navigate to Flows → Create flow
- 4 Add a condition node that checks if the input contains 'summarize' and routes to the appropriate prompt
- 5 Test the flow with both types of queries and verify correct routing
Practice Advanced Prompting Patterns
Implement chain-of-thought, few-shot, and JSON schema enforcement in the Bedrock playground.
- 1 Open Bedrock playground and test a zero-shot prompt: 'Is 17 a prime number?'
- 2 Add chain-of-thought: 'Think step by step. Is 17 a prime number?'
- 3 Create a few-shot prompt with 2 examples of the desired output format before the actual question
- 4 Add JSON schema enforcement: 'Respond in this exact JSON format: {"answer": boolean, "reasoning": string}'
- 5 Compare output quality and consistency across all 4 approaches
Scenarios
Think through each scenario before revealing the answer.
Centralized Prompt Management
- •Which Bedrock feature centralizes prompt templates?
- •How do you version and track changes?
- •Which AWS service provides audit trails for API calls?
Practice Questions
15 questions across 3 difficulty levels.
Further Reading
Go deeper into today's topics.
Prompt Evaluation with Prompt Management and Flows (Notebook)
Hands-on notebook: create prompt variants, build evaluation flows, compare outputs programmatically.
Guide to Prompt Optimization in Amazon Bedrock
Auto-rewrite prompts for better accuracy: Prompt Optimization feature reduces trial-and-error.
Implementing Advanced Prompt Engineering with Bedrock
Chain-of-thought, few-shot, role-play, and structured output techniques with Bedrock models.
Getting Started with Prompt Management Flows
Step-by-step: build prompt chains with Flows visual builder, integrate KB retrieval nodes.
Implement Prompt Caching on Bedrock — Cut Costs in Half
Practical implementation: cache static context, automatic vs manual checkpoints, 90% input token savings.