Unit 6: Data Analysis & Visualization; AI Model Environments & Lifecycle Basics - Practice Quiz

1 Which AI tool allows users to upload files and perform data analysis using natural language prompts?

2 Tableau is primarily used for which of the following tasks?

3 What is the main benefit of using data visualization tools like Tableau?

4 Which of the following is an example of structured data?

5 Which of the following is considered unstructured data?

6 Structured data is typically stored in which format?

7 What is the primary purpose of a data pipeline?

8 In the ETL process, what does the letter 'T' stand for?

9 Why is automation useful in data pipelines?

10 Which of the following is a benefit of using cloud services for AI?

11 Which of the following is a well-known cloud service provider?

12 What does the term 'pay-as-you-go' mean in cloud computing?

13 What does edge deployment mean in AI?

14 Which is a key advantage of edge deployment over cloud-only processing?

15 Which of the following is an example of an edge device?

16 What is the first step in fixing a problem in a data workflow?

17 What is troubleshooting in the context of AI systems?

18 What does AI process automation aim to achieve?

19 What does MLOps primarily focus on?

20 Which stage is part of the machine learning model lifecycle?

21 A student uploads a CSV of monthly sales to ChatGPT Advanced Data Analysis and asks for a trend chart. Which capability makes this possible?

22 In Tableau, you want to compare sales across regions and see the magnitude at a glance using color intensity on a map. Which visualization is most appropriate?

23 Which of the following is the best example of unstructured data?

24 A team must analyze thousands of scanned invoices to extract totals into a database. Which preprocessing step is essential to convert this unstructured input into structured data?

25 In an ETL pipeline, which stage is responsible for cleaning and reshaping raw data before it is stored?

26 A company's data pipeline reprocesses the entire dataset every night even when only a few rows change. Which improvement best reduces wasted compute?

27 A startup wants to run ML training without managing servers, paying only for the compute time used per function call. Which cloud model fits best?

28 Which cloud service model gives a data science team the most control over the operating system and installed libraries while still avoiding physical hardware management?

29 A smart camera must detect defects on a factory line with sub-second latency and no internet dependency. Why is edge deployment preferred over cloud inference here?

30 Which technique is most commonly used to make a large neural network fit and run efficiently on a resource-constrained edge device?

31 A model shows 99% accuracy on training data but 62% on test data. What error is most likely occurring?

32 While analyzing a dataset, you notice a column of ages contains the value and another value . What type of data issue is this?

33 A Python data script fails with a KeyError when accessing a column named 'Sales'. What is the most effective first troubleshooting step?

34 A deployed model's predictions have gradually become less accurate over several months, though the code is unchanged. What should you investigate first?

35 An insurance firm wants to automate reading emailed claim forms and routing them to the right department. Which combination best describes this AI-driven automation?

36 Which scenario is the LEAST suitable candidate for AI process automation?

37 In the ML lifecycle, which MLOps practice ensures you can reproduce exactly which dataset and code produced a given model?

38 Why is continuous monitoring considered essential after a model is deployed to production?

39 A team wants every code commit to automatically train, test, and validate a model before deployment. Which MLOps concept does this describe?

40 You want to show the relationship between advertising spend and revenue and check whether higher spend correlates with higher revenue. Which chart type is most appropriate?

41 You upload a 2 GB CSV to ChatGPT Advanced Data Analysis and ask it to compute a rolling 30-day average grouped by region. The tool returns partial results and warns about memory. What is the most robust fix within the tool's execution sandbox?

42 In Tableau, a LOD expression {FIXED [Region] : SUM([Sales])} produces different values than a table calculation of the same aggregation. Why?

43 A pipeline ingests customer emails (unstructured) and CRM records (structured). To join them for churn analysis, which approach best bridges the modality gap?

44 A data lake stores JSON logs with inconsistent schemas across days. A query engine using schema-on-read fails on new fields. What is the correct interpretation?

45 An ETL DAG has a task that occasionally succeeds after retry due to a flaky upstream API. To keep the pipeline both reliable and idempotent, which combination is best?

46 In a streaming pipeline, you must guarantee that each event affects downstream aggregates exactly once even if the source redelivers messages. Which mechanism directly enables this?

47 A model inference service on the cloud has spiky, unpredictable traffic and strict cost limits during idle periods. Which deployment model minimizes idle cost while handling spikes?

48 You notice cross-region data transfer charges dominating your cloud bill for an analytics pipeline. The compute is in region A but the data lake is in region B. What is the most cost-effective architectural change?

49 You must deploy a vision model to battery-powered edge devices with 512 MB RAM and no reliable network. Which optimization set best fits these constraints?

50 An edge fleet runs a quantized model, but field accuracy is lower than in the lab. Logs show input images have different lighting than training data. What is the root cause and fix?

51 A model reports 99% training accuracy but 62% validation accuracy, and validation loss rises while training loss keeps falling. What does this pattern indicate?

52 During analysis you find a feature perfectly predicts the target in training but is unavailable at prediction time in production. This is an example of:

53 A batch scoring job intermittently produces NaN predictions for a small subset of records. The model and code are unchanged. What is the most systematic first troubleshooting step?

54 After a deployment, inference latency at p99 spiked while p50 stayed flat. Throughput is unchanged. What does this most likely indicate?

55 You automate an approval workflow with an AI classifier that routes low-risk cases automatically and escalates uncertain ones. Which design best balances automation and safety?

56 An RPA bot that scrapes a web portal breaks whenever the site's HTML layout changes. Which strategy most durably reduces this fragility?

57 Your production model's accuracy slowly degrades over months even though the code is frozen. Monitoring shows input feature distributions drifting from training. Which MLOps practice directly addresses this?

58 A team wants to safely release a new model version while limiting blast radius if it underperforms. Which deployment strategy fits best?

59 For reproducible ML experiments, which combination of artifacts must be versioned together to guarantee a result can be regenerated exactly?

60 A stakeholder claims a Tableau dashboard shows sales "trending up," but the line uses a dual axis with two independently scaled measures. Why is this visualization misleading?