GK Questions

Data minimization (DPDP Act, GDPR) limits collection to what's necessary for specified purpose. Reduces privacy risks, storage costs, breach impact. Critical for responsible AI data practices.

Proportion visualization: bar charts for clear comparison, pie charts for part-to-whole (limited categories), stacked bars for multiple groups. Choice depends on data and audience.

Scatter plots display individual data points for two variables, revealing correlations, clusters, outliers. Foundation for regression analysis and exploratory data science.

Bias in training data or model design can lead to discriminatory outcomes. Mitigation: diverse datasets, fairness metrics, bias audits, human oversight. Critical for ethical AI deployment.

Transparency (Explainable AI/XAI) ensures AI decisions can be interpreted by humans. Techniques: LIME, SHAP, attention visualization. Critical for trust, debugging, regulatory compliance.

Data integration combines disparate sources via ETL/ELT, virtualization, or federation. Enables unified analytics, reporting, and decision-making. Critical for enterprise data strategy.

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.

Serverless query engines (AWS Athena, Google BigQuery) enable SQL queries directly on data in object storage using schema-on-read. Eliminates ETL for ad-hoc analysis. Pay-per-query pricing.

IaC (Terraform), Configuration Management (Ansible), GitOps all use version-controlled code to define infrastructure state. Enables repeatability, auditability, peer review, automated testing.

Data lakes (S3, ADLS) store raw structured/unstructured data for flexible analysis; data warehouses (Snowflake, Redshift) store curated, schema-on-write data for BI. Modern: lakehouse pattern combines both.

Managed Kubernetes: EKS (AWS), AKS (Azure), GKE (GCP) handle control plane, updates, scaling. Abstracts infrastructure while retaining container flexibility. Critical for cloud-native development.

Parquet is columnar storage format optimized for analytics: efficient compression, predicate pushdown, schema evolution. Reduces I/O by reading only required columns. Used in Spark, Hive, Presto.

Serverless (Lambda, Cloud Functions) auto-scales based on demand, charges per execution. Trade-offs: cold starts, vendor lock-in, execution limits. Critical for cost-optimized cloud design.

Resilience patterns: self-healing replaces failed instances, circuit breaker prevents cascading failures, retry logic handles transient errors. Combined for fault-tolerant cloud applications.

OECD AI Principles (2019) provide values-based guidelines: inclusive growth, human-centered values, transparency, robustness, accountability. India endorses for domestic policy alignment.

National Cyber Security Strategy (draft) addresses: critical infrastructure protection, cyber diplomacy, capacity building, public-private partnerships. Foundation for India's cyber defense posture.

Data Protection Board of India (under DPDP Act 2023) adjudicates disputes, imposes penalties, enforces compliance. Distinct from CERT-In (cyber incidents) and MeitY (policy).

MeitY's FOSS policy encourages government adoption of open-source for cost-effectiveness, security, and vendor independence. Critical for sustainable digital governance.

High-dimensional handling: dimensionality reduction (PCA, autoencoders) compresses features; sparse representations (L1 regularization) select relevant features. Critical for text, recommendation systems.

Feature scaling (standardization, normalization) prevents features with large ranges from dominating model learning. Critical for distance-based algorithms and gradient descent optimization.