Create a custom practice set
Pick category, difficulty, number of questions, and time limit. Start instantly with your own quiz.
Generate QuizPick category, difficulty, number of questions, and time limit. Start instantly with your own quiz.
Generate QuizNo weekly quiz is published yet. Check the weekly page for the latest updates.
View Weekly PageFilter by category, type, and difficulty. Reading is open for everyone.
Answer: Both A and B
Digital Public Goods Alliance curates open-source solutions; G20 DPI Framework promotes interoperable identity, payments, data sharing. India leads global DPI advocacy.
Answer: Model-Based Imputation
Model-based imputation (MICE, Bayesian) models relationships between variables to predict missing values, accounting for uncertainty. More robust than simple methods for complex missingness patterns.
Answer: Event-Driven Architecture
Event-driven architecture uses message brokers (Kafka, RabbitMQ) for decoupled, asynchronous service communication. Enables scalability, resilience, and real-time processing.
Answer: Reinforcement / RLHF
Reinforcement Learning from Human Feedback (RLHF) trains reward models from human preferences, then optimizes LLM to maximize rewards. Critical for aligning AI with human values.
Answer: All of these
Effective tech policy: proportionate to actual risks, technology-neutral (focus on outcomes), risk-based (higher scrutiny for higher-risk activities). Critical for balanced policy design.
Answer: All of these
Imbalance handling: oversampling (SMOTE) increases minority, undersampling reduces majority, class weighting adjusts loss function. Choice depends on data size and model type.
Answer: Both A and B
Stateless microservices store session data externally (Redis, database), enabling horizontal scaling by adding instances. Critical for elastic, resilient cloud applications.
Answer: Beam / Nucleus
Advanced decoding: beam search explores multiple hypotheses, nucleus (top-p) sampling truncates low-probability tokens. Balance diversity and coherence in text generation.
Answer: Both A and B
India endorses OECD AI Principles (2019) and UNESCO Recommendation on AI Ethics (2021). Both emphasize human rights, transparency, accountability, sustainability. Critical for AI governance.
Answer: All of these
Time series methods: ARIMA for stationary series, Prophet for seasonality/holidays, LSTM for complex patterns. Choice depends on data characteristics and forecast horizon.
Answer: All of these
Safe deployment patterns: blue-green (instant switch), canary (gradual rollout), feature flags (toggle functionality). Combined for zero-downtime, low-risk releases. Critical for DevOps practices.
Answer: Multimodal Learning
Multimodal models (CLIP, LLaVA) process text, images, audio jointly, enabling visual question answering, image captioning, and cross-modal retrieval. Critical for next-gen AI applications.
Answer: All of these
High-cardinality handling: one-hot for low cardinality, target encoding for medium, embeddings for high (deep learning). Choice depends on model type and cardinality. Critical for feature preprocessing.
Answer: Multi-Site Active-Active
Active-active runs full workload in multiple regions simultaneously, enabling instant failover with zero RPO/RTO. Highest cost but maximum resilience. Critical for mission-critical systems.
Answer: AI / Autonomous
AI agents combine LLMs with planning, tool use, memory for autonomous task execution. Applications: research assistants, customer service, workflow automation. Critical for next-gen AI systems.
Answer: Both A and B
ISO/IEC 24760 provides identity management framework; W3C DID enables decentralized identifiers. Both support interoperable, privacy-preserving digital identity systems. Critical for identity policy questions.
Answer: SMOTE
SMOTE (Synthetic Minority Over-sampling Technique) creates synthetic examples by interpolating between minority class neighbors. Improves classifier performance on imbalanced data vs simple oversampling.
Answer: Event-Driven Architecture
Event-driven architecture uses message brokers (Kafka, RabbitMQ) for asynchronous, decoupled service communication. Enables scalability, resilience, and real-time processing. Critical for modern distributed systems.
Answer: All of these
Malware analysis combines: static (code inspection), dynamic (runtime behavior), sandboxing (isolated execution). Critical for threat intelligence and incident response.
Answer: Pruning
Model pruning removes low-importance weights/neurons, reducing size and inference cost. Combined with quantization for efficient edge deployment. Critical for scalable AI systems.