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.