Technology – GK Questions

Histograms bin continuous data to show frequency distribution, revealing skewness, outliers, and modality. Critical for exploratory data analysis and statistical understanding.

IaC (Terraform, CloudFormation) defines infrastructure in declarative code, enabling version control, peer review, automated testing, and consistent deployments. Critical for cloud governance.

Refactor (re-architect) redesigns applications for cloud-native: microservices, serverless, containers. Maximizes cloud benefits but requires significant effort. Critical for digital transformation.

Hybrid cloud combines on-premises infrastructure with public cloud services. Enables data sovereignty compliance while leveraging cloud scalability. Critical for regulated industries.

Multi-cloud architecture distributes workloads across AWS, Azure, GCP to avoid vendor lock-in, improve resilience, and leverage best-of-breed services. Critical for enterprise cloud strategy.

SOAR (Security Orchestration, Automation, and Response) automates playbooks: alert triage, enrichment, containment actions. Reduces analyst workload and response time. Critical for SOC efficiency.

NIST SP 800-61 and ISO 27035 provide structured incident response methodologies. Critical for consistent, effective handling of security incidents in enterprises.

Threat hunting uses hypotheses, analytics, and analyst expertise to find hidden adversaries. Complements automated detection. Critical for advanced persistent threat (APT) defense.

Threat intelligence (commercial, open-source, ISACs) provides IOCs: malicious IPs, domains, hashes, TTPs. Enables proactive blocking and hunting. Critical for mature security operations.

SIEM (Security Information and Event Management) aggregates logs, correlates events, and detects anomalies across network, endpoint, and application sources. Foundation for modern SOC operations.

Multi-task learning shares representations across related tasks, improving sample efficiency and generalization. Critical for building versatile AI systems with limited labeled data per task.

MLOps platforms integrate: feature stores for consistent data, model registries for versioning, serving infrastructure for scalable inference. Critical for production ML system reliability.

Quantization (INT8, INT4) reduces model size 2-4x with minimal accuracy loss, enabling deployment on edge devices and reducing cloud inference costs. Critical for scalable LLM serving.

Function calling/tool use enables LLMs to invoke APIs, run code, query databases for tasks beyond text generation. Critical for building agentic AI systems that interact with real-world systems.

Artifact management (MLflow, DVC) tracks datasets, model versions, hyperparameters, and code commits. Enables reproducibility, auditability, and rollback in ML workflows. Critical for enterprise MLOps.

MLOps combines: CI/CD for automated deployment, drift detection for performance monitoring, A/B testing for validation. Critical for maintaining ML system reliability and business value over time.

PEFT methods (LoRA, Adapter, Prefix-tuning) update small subset of parameters or add lightweight modules, reducing compute/memory vs full fine-tuning. Critical for resource-constrained LLM adaptation.

Retrieval-Augmented Generation (RAG) combines LLMs with external knowledge retrieval, reducing hallucinations by grounding responses in verified sources. Critical for factual accuracy in enterprise AI applications.

Chain-of-Thought prompting encourages LLMs to generate reasoning steps before final answer, improving performance on complex reasoning tasks. Critical for enhancing LLM reliability in high-stakes applications.

Fine-tuning updates pre-trained model weights on task-specific data, enabling adaptation to domains like healthcare, legal, or customer service with limited labeled examples. Critical for practical LLM deployment.