Prepare Sample Input and Prompt

Sample Conversation

Use a short transcript first. This reduces risk and helps validate the Bedrock analysis before adding audio transcription.

Example:

Mentor: Why did you choose this project?
Student: I want to build an AI assistant for communication.
Mentor: Why is that useful?
Student: Because many people cannot explain ideas clearly under pressure.
Mentor: Why should this use AWS?
Student: AWS provides storage, transcription, AI analysis, workflow orchestration, and monitoring.
Mentor: What happens if AI is wrong?
Student: The system should be used as coaching feedback, not final truth.

Bedrock Prompt Structure

The prompt should ask the model to return a structured coaching report:

You are a communication coach. Analyze the transcript.
Return:
1. Short summary
2. Main topic
3. Strong points
4. Weak reasoning points
5. Improved answer using claim, reason, evidence, example
6. Five why-chain practice questions
7. Safety note that feedback is a suggestion
8. Vietnamese summary

Expected Output Quality

The report should be:

  • Specific to the transcript.
  • Practical and not generic.
  • Framed as coaching feedback.
  • Bilingual where required.
  • Clear enough for a learner to practice from.

Validation

Before automating the full workflow, manually test the prompt in the Amazon Bedrock console with the sample transcript and save a screenshot of the generated report.