Technology – Medium Level – GK Questions

MLOps combines CI/CD, drift detection, A/B testing for production ML.

Function calling enables LLMs to invoke APIs, run code, query databases.

PEFT methods update small subset of parameters, reducing compute/memory.

Chain-of-Thought prompting encourages LLMs to generate reasoning steps.

Clustering discovers natural groupings in unlabeled data.

Feature engineering: domain knowledge, transformations, interactions, encoding.

Time series forecasting models temporal patterns for demand prediction, anomaly detection.

Serverless query engines enable SQL queries on object storage using schema-on-read.

Data lakes store raw structured/unstructured data; warehouses store curated data.

Parquet is columnar storage format optimized for analytics: compression, predicate pushdown.

Stream processing handles continuous data flows for real-time insights.

CDC captures database operations and streams to targets for real-time analytics.

IOCs are forensic artifacts indicating potential breaches. Used for proactive blocking.

Micro-segmentation creates granular security zones, containing breaches and limiting attacker movement.

MFA requires multiple authentication factors; adaptive authentication adjusts requirements based on risk context.

FAME provides demand incentives for EVs and charging infrastructure.

Positional encodings add sequence order information to token embeddings, critical for language understanding tasks.

Self-attention computes weighted relationships between tokens, enabling context-aware processing. Foundation of Transformer success.

Data catalogs enable discovery; data lineage tracks data flow and transformations. Both critical for data governance.

Edge computing processes data near source, reducing round-trip latency to cloud. Critical for IoT, autonomous vehicles.