GEN AI TECH QUIZ - 03/03/2026

A retail company wants to forecast next month’s sales using historical structured data stored in Excel sheets. They do not need text genera on or chatbot func onality. As an AI engineer, which approach would you recommend?

• Large Language Model

• Genera ve AI Agent

• Tradi onal ML Regression Model

• Mul-agent orchestra on

You built a chatbot using an LLM API. The client complains that for the same ques on, the answer changes every me. They want more consistent and reproducible responses. What should you adjust first?

• Increase temperature

• Reduce temperature close to 0

• Increase top-p

• Increase max tokens

A hiring assistant trained on historical company data starts favoring one demographic group unfairly. The management wants the system to be more responsible and unbiased. What is the most appropriate ac on?

• Increase token limit

• Audit dataset and retrain with balanced data

• Add more agents

• Increase inference speed

During discussion, a stakeholder says they want “AI” but their actual requirement is classifica on of customer churn. There is no need for content genera on. Which technology is most suitable?

• LLM with RAG

• Genera ve AI model

• Supervised Machine Learning

• Mul-agent workflow

In the LLM lifecycle, a er pretraining on massive internet data, the model is refined for specific behavior like instruc on-following. This stage improves alignment with user tasks. What is this stage called?

• Inference

• Fine-tuning

• Deployment

• Chunking

You are integrating OpenAI API using Python. The response object is returned successfully. You want to access structured JSON data from it. Which method should be used?

• response.text

• response.json()

• response.data()

• json.parse(response)

While tes ng your API integra on, you receive HTTP status code 401. The API endpoint exists and internet is working fine. What does this error most likely indicate?

• Internal server failure

• Resource not found

• Unauthorized access (invalid API key)

• Successful response

You are building a unified LLM wrapper that connects to OpenAI, Bedrock, and Gemini. Different APIs return responses in different formats. What is the best design principle to follow?

• Hardcode format for each provider

• Normalize responses into a common structure

• Remove logging

• Ignore error handling

Your teammate hardcoded API keys directly into a Python file and pushed it to GitHub. From a produc on and security standpoint, what is the correct approach instead?

• Keep keys in code

• Store in .env file and load securely

• Put inside prompt

• Share via email

When sending prompts to LLM APIs, large text input is usually sent in request body. Which HTTP method is most appropriate for sending structured data in body format?

• GET

• POST

• DELETE

• OPTIONS