Download Future certifications Practice Exam and more Exams Technology in PDF only on Docsity!
Question 1. Which learning paradigm involves an agent learning to make decisions by receiving rewards or penalties for its actions? A) Supervised learning B) Unsupervised learning C) Reinforcement learning D) Semi‑supervised learning Answer: C Explanation: Reinforcement learning trains agents through interaction with an environment, using reward signals to shape behavior. Question 2. In the context of AI, what does the term “transformer” refer to? A) A type of recurrent neural network B) A model architecture that relies on self‑attention mechanisms C) A hardware accelerator for deep learning D) A preprocessing technique for image data Answer: B Explanation: Transformers use self‑attention to weigh relationships between all tokens, enabling efficient parallel processing of sequences. Question 3. Which of the following is NOT a typical step in the Machine Learning lifecycle? A) Data ingestion B) Model training C) Model decommissioning D) Model monitoring Answer: C Explanation: Decommissioning is not a standard lifecycle step; after deployment, models are monitored, updated, or retired, but “decommissioning” is not a distinct phase.
Question 4. In supervised learning, the term “label” refers to: A) The input features of a dataset B) The target variable the model tries to predict C) The algorithm used for training D) The evaluation metric Answer: B Explanation: Labels are the ground‑truth outputs associated with each training example. Question 5. Which metric is most appropriate for evaluating a binary classification model when the classes are highly imbalanced? A) Accuracy B) Precision C) Recall D) Area Under the ROC Curve (AUC) Answer: D Explanation: AUC summarizes performance across all thresholds and is less affected by class imbalance than accuracy. Question 6. What is the primary purpose of dimensionality reduction techniques such as PCA? A) Increase the number of features B) Remove outliers from the dataset C) Transform correlated variables into a smaller set of uncorrelated components D) Normalize feature values to a common scale Answer: C
D) A convolutional encoder Answer: B Explanation: RAG retrieves relevant documents from an external store and feeds them to the LLM to improve factual accuracy. Question 10. Which of the following is a common method for handling missing values in a dataset? A) One‑hot encoding B) Mean imputation C) Feature scaling D) Gradient clipping Answer: B Explanation: Mean imputation fills missing numeric entries with the average value of that feature. Question 11. In feature scaling, the difference between standardization and normalization is that: A) Standardization centers data to zero mean and unit variance; normalization rescales to a fixed range (e.g., 0‑1) B) Normalization centers data; standardization rescales to 0‑ 1 C) Both are identical processes D) Standardization only works on categorical data Answer: A Explanation: Standardization uses mean and standard deviation, while normalization applies min‑max scaling. Question 12. Which algorithm is best suited for clustering unlabeled data? A) Linear regression
B) K‑Means C) Decision tree D) Logistic regression Answer: B Explanation: K‑Means partitions data into k clusters based on distance metrics without requiring labels. Question 13. What does “overfitting” imply about a model’s performance? A) It performs well on both training and unseen data B) It has high bias and low variance C) It captures noise in the training data, leading to poor generalization D) It underutilizes available features Answer: C Explanation: Overfitting occurs when a model learns the training data too closely, reducing its ability to generalize. Question 14. Which MLOps practice helps detect model drift after deployment? A) A/B testing B) Continuous integration C) Real‑time performance monitoring against baseline metrics D) Model compression Answer: C Explanation: Monitoring key performance indicators over time reveals drift when the model’s predictions deviate from expected behavior. Question 15. Docker containers are isolated using: A) Virtual machines
C) Amazon Redshift D) Amazon Aurora Answer: B Explanation: DynamoDB provides key‑value and document storage without server management. Question 19. In Azure, which resource type implements a virtual network isolation boundary? A) Resource Group B) Virtual Network (VNet) C) Availability Set D) Storage Account Answer: B Explanation: A VNet defines IP address space, subnets, and routing for isolated network communication. Question 20. The Shared Responsibility Model in cloud computing states that: A) The cloud provider is responsible for all security controls B) The customer is responsible for hardware maintenance C) Security responsibilities are divided between provider and customer depending on the service model D) Only the provider handles compliance Answer: C Explanation: The model delineates which security tasks belong to the provider (e.g., physical security) and which to the customer (e.g., data encryption). Question 21. Which of the following is a typical use case for serverless functions (e.g., AWS Lambda, Azure Functions)?
A) Running a continuously active web server B) Processing event‑driven workloads such as image resizing on upload C) Hosting a relational database D) Managing a Kubernetes cluster Answer: B Explanation: Serverless functions excel at short‑lived, event‑triggered tasks. Question 22. What is the primary advantage of using reserved instances in cloud cost optimization? A) Unlimited scaling B) Lower hourly rates in exchange for a commitment to a specific instance type and region C) Automatic failover across zones D) Ability to run on-premises workloads Answer: B Explanation: Reserved instances provide a discounted rate for a pre‑committed usage period. Question 23. Which IAM principle follows the “least privilege” concept? A) Granting admin rights to all users B) Assigning permissions only necessary for a user’s job function C) Using a single shared root account D) Allowing all actions on all resources by default Answer: B Explanation: Least privilege limits access to the minimum required to perform duties. Question 24. Data encryption “at rest” protects data when: A) It is being transmitted over a network
C) Reliance on perimeter firewalls only D) No encryption required for internal traffic Answer: B Explanation: Zero Trust assumes no implicit trust and enforces strict identity and device verification. Question 28. IPsec is primarily used to: A) Secure email transmission B) Encrypt web traffic C) Provide secure VPN tunnels at the network layer D) Authenticate users via passwords Answer: C Explanation: IPsec secures IP packets, enabling site‑to‑site or remote‑access VPNs. Question 29. Which wireless security protocol is the most recent and secure for Wi‑Fi networks? A) WEP B) WPA C) WPA D) WPA Answer: D Explanation: WPA3 introduces stronger encryption and protection against dictionary attacks. Question 30. A Security Information and Event Management (SIEM) system is used for: A) Automating software builds B) Aggregating, correlating, and analyzing security logs in real time C) Managing virtual machine snapshots
D) Deploying container images Answer: B Explanation: SIEM collects logs from multiple sources, applies correlation rules, and alerts on suspicious activity. Question 31. In the Incident Response lifecycle, the “containment” phase aims to: A) Identify the root cause of the breach B) Eradicate malicious artifacts C) Limit the spread of the incident while preserving evidence D. Recover normal operations Answer: C Explanation: Containment isolates affected systems to prevent further damage while maintaining forensic data. Question 32. Which backup strategy involves taking a full backup periodically and incremental backups in between? A) Grandfather‑Father‑Son B) Full + Incremental C) Differential only D) Continuous Data Protection (CDP) Answer: B Explanation: Full backups capture all data; incremental backups record only changes since the last backup. Question 33. Structured Query Language (SQL) “JOIN” operations are used to: A) Combine rows from two or more tables based on related columns B) Encrypt database tables
D. You need a probability‑based metric Answer: B Explanation: MAE averages absolute errors, providing a linear penalty and being robust to outliers compared to RMSE. Question 37. In Python’s Pandas library, which function is used to remove duplicate rows? A) dropna() B) rename() C) drop_duplicates() D. pivot_table() Answer: C Explanation: drop_duplicates() eliminates rows that have identical values across selected columns. Question 38. Which visualization is most appropriate for showing the distribution of a single continuous variable? A) Bar chart B) Scatter plot C) Histogram D. Pie chart Answer: C Explanation: Histograms bin continuous data to show frequency distribution. Question 39. A “dashboard” in a Business Intelligence tool typically provides: A) Raw data files for download only B) Real‑time or near‑real‑time visual summaries of key metrics C. Source code for data pipelines
D. Automated machine learning model training Answer: B Explanation: Dashboards aggregate and display important KPIs in an easily digestible visual format. Question 40. In Agile Scrum, which artifact represents the work to be completed in a sprint? A) Product backlog B) Sprint backlog C) Burndown chart D. Definition of Done Answer: B Explanation: The sprint backlog lists tasks selected from the product backlog for the current sprint. Question 41. Which DevOps practice emphasizes “Infrastructure as Code” (IaC)? A) Manual server provisioning B) Writing configuration scripts in Terraform, CloudFormation, or similar tools C) Using spreadsheets to track resources D. Hard‑coding IP addresses in application code Answer: B Explanation: IaC treats infrastructure definitions as version‑controlled code, enabling automated provisioning. Question 42. In a CI/CD pipeline, which stage is responsible for automatically building the application after code is committed? A) Deployment
A) CPU utilization B) Network latency C) Heap memory usage over time D. Disk I/O throughput Answer: C Explanation: A steadily increasing heap size indicates that memory is not being reclaimed, suggesting a leak. Question 46. In the context of SRE, an “error budget” is: A) The total amount of downtime a service can tolerate while still meeting its SLO B) The maximum number of bugs allowed in a release C. The budget allocated for purchasing monitoring tools D. The amount of money saved by automating tests Answer: A Explanation: Error budget quantifies permissible unreliability, balancing reliability goals with feature velocity. Question 47. Which cloud service is a managed data warehouse optimized for analytics? A) Amazon S B) Amazon Redshift C) Amazon EC D. Amazon SQS Answer: B Explanation: Redshift stores columnar data and provides fast query performance for analytical workloads. Question 48. What does “stateless” mean for a microservice?
A) It retains user session data locally B) It does not store any client‑specific data between requests, relying on external storage if needed C. It cannot scale horizontally D. It runs without any configuration files Answer: B Explanation: Stateless services treat each request independently, simplifying scaling and resilience. Question 49. Which of the following is a common technique for preventing “SQL injection” attacks? A) Using SELECT * statements B. Storing passwords in plain text C) Parameterized queries or prepared statements D. Disabling input validation Answer: C Explanation: Parameterized queries separate code from data, neutralizing malicious input. Question 50. In a VPC, a “subnet” is primarily used to: A) Define a security group B) Partition the IP address range into smaller, isolated networks C. Provide DNS resolution D. Manage IAM roles Answer: B Explanation: Subnets allocate subsets of the VPC’s CIDR block for logical grouping and isolation.
Question 54. Which evaluation metric is appropriate for a regression model predicting house prices? A) Accuracy B) F1‑Score C) Root Mean Squared Error (RMSE) D. Precision Answer: C Explanation: RMSE measures the average magnitude of prediction errors in the same units as the target. Question 55. What is the purpose of “cross‑validation” in model training? A) To increase the size of the training set by duplication B) To assess model performance on multiple train‑test splits, reducing variance in evaluation C. To automatically select hyperparameters without human input D. To encrypt the training data Answer: B Explanation: Cross‑validation rotates validation folds, providing a more reliable estimate of generalization. Question 56. In the context of cloud networking, an “Internet Gateway” provides: A) Private connectivity between VPCs B) A route for traffic from the VPC to the public internet C. Encrypted storage for logs D. Auto‑scaling for compute instances Answer: B Explanation: An Internet Gateway enables inbound and outbound internet traffic for resources in a VPC.
Question 57. Which GCP service is a fully managed, serverless data warehouse? A) Cloud Storage B) BigQuery C) Compute Engine D. Cloud SQL Answer: B Explanation: BigQuery allows SQL‑like analytics on massive datasets without infrastructure management. Question 58. Which principle ensures that security controls are applied consistently across multiple cloud providers? A) Vendor lock‑in B) Multi‑cloud governance C. Data residency D. Elastic scaling Answer: B Explanation: Multi‑cloud governance defines policies and controls that span different cloud platforms. Question 59. In a CI pipeline, “static code analysis” is used to: A) Execute unit tests on compiled code B. Identify code quality issues, security vulnerabilities, and style violations without running the program C. Deploy the application to production D. Monitor runtime performance metrics Answer: B
