GoofyCubes
25

Chapter 25

Final Quiz

Learning Objective

Test your understanding of GenAI project implementation concepts.

Multiple Choice

1. What is the primary purpose of RAG?

  • A. Train a new LLM
  • B. Retrieve trusted context before generating an answer
  • C. Reduce Docker image size
  • D. Replace APIs

RAG retrieves relevant enterprise knowledge and gives it to the LLM so the response is grounded in trusted documents.

2. Which tool is best suited for stateful multi-agent workflows?

  • A. LangGraph
  • B. GitHub
  • C. Dockerfile
  • D. SQL only

LangGraph is designed for stateful, graph-based agent workflows with branching, retries, and human review.

3. What is a Docker image?

  • A. A running pod
  • B. A blueprint used to create containers
  • C. A vector database
  • D. A prompt template

A Docker image is the blueprint; a container is the running instance of that image.

4. What is a Kubernetes pod?

  • A. A code repository
  • B. The smallest deployable unit in Kubernetes
  • C. An LLM parameter
  • D. A Git branch

A pod is the smallest deployable unit in Kubernetes and usually runs one application container.

5. What is prompt injection?

  • A. Adding embeddings to a vector store
  • B. A malicious attempt to override model instructions
  • C. A CI/CD deployment method
  • D. A token counter

Prompt injection tries to manipulate the LLM into ignoring its original instructions or revealing sensitive data.

Scenario Quiz

1. A model is answering from memory instead of policy documents. What should you do?

Answer: Use grounded RAG, require citations, and validate that answers are based on retrieved sources.

2. A prompt is producing inconsistent JSON. What should you do?

Answer: Strengthen the prompt, define schema clearly, use low temperature, validate JSON, and retry or route failures.

3. A GenAI API works locally but fails in QA due to dependency mismatch. What helps?

Answer: Docker packaging with pinned dependencies.

4. Traffic increases and the GenAI API becomes slow. What helps?

Answer: Kubernetes scaling, caching, model routing, and token optimization.

5. Reviewers complain that summaries miss important details. What should be checked?

Answer: Prompt design, document parsing, retrieval quality, chunking, evaluation dataset, and user feedback logs.

Architect Takeaway

A production GenAI project is a system of controls: architecture, data quality, retrieval, prompts, models, validation, security, DevOps, deployment, monitoring, and continuous improvement.