GK Questions

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.

Transformer architecture (Vaswani et al., 2017) uses self-attention mechanisms to process sequences in parallel, enabling scaling to billions of parameters. Foundation for GPT, BERT, LLaMA, and other LLMs. Critical for understanding modern AI systems.

Compiling source code is performed by compiler software (like GCC, javac), not the OS. OS functions include: process management, memory management, file system management, device management, security, and user interface. Compilers are application/system software that run ON the OS.

QR (Quick Response) Code is a two-dimensional barcode that stores payment information. UPI apps can scan merchant QR codes to initiate payments instantly without entering account details. NPCI's Bharat QR is interoperable across banks and payment apps, promoting cashless transactions at small merchants.

Computer Vision enables machines to interpret and understand visual information from the world using cameras and AI algorithms. Applications include facial recognition, medical imaging analysis, autonomous vehicles, quality inspection, and augmented reality. Combines deep learning, image processing, and pattern recognition.

Internet is the global network of interconnected computers using TCP/IP protocols. WWW (World Wide Web) is a service that runs on the internet, using HTTP/HTTPS to access interlinked hypertext documents. Other internet services include email, FTP, VoIP. Web is a subset of internet.

VSSC (Vikram Sarabhai Space Centre), Thiruvananthapuram, is ISRO's lead center for launch vehicle development including PSLV, GSLV, and LVM3. URSC handles satellites, LPSC develops propulsion systems, ISAC integrates satellites. ISRO has multiple specialized centers across India.

Hashing transforms input data of any size into a fixed-length hash value using algorithms like SHA-256 or MD5. It is one-way (irreversible), used for password storage, data integrity verification, and digital signatures. Unlike encryption, hashed data cannot be decrypted to original form.