Glossary

Agentic AI refers to AI systems that autonomously plan, decide, and execute multi-step actions to achieve a goal, typically by combining a language model with tools, memory, and feedback loops.

Retrieval-Augmented Generation (RAG) is a pattern where an AI system retrieves relevant documents from a knowledge source and includes them in the prompt, so the model answers based on grounded, current information rather than training-time memory alone.

An LLM application is software built around one or more large language models that interact with users or systems, typically combining prompts, tools, retrieval, evaluation, and traditional code to deliver a specific business outcome.

MLOps is the practice of operating machine learning systems in production reliably and repeatably — covering training, deployment, monitoring, retraining, and governance, analogous to DevOps for software.

A vector database is a data store optimized for storing and searching high-dimensional vector embeddings, typically used to find semantically similar items for AI applications such as RAG, recommendation, and search.

Managed Detection and Response (MDR) is a security service that combines 24x7 monitoring, threat hunting, AI-assisted triage, and incident response — typically delivered by a third-party SOC on top of customer-owned telemetry.

AI governance is the set of policies, processes, and controls organizations use to develop, deploy, and operate AI systems responsibly — covering risk management, model lifecycle, accountability, data protection, and regulatory compliance.

Observability is the ability to understand a system’s internal state by examining its external outputs — metrics, logs, traces, and (increasingly) profiles — without needing to ship new code to investigate a problem.

GitOps is an operational pattern where the desired state of infrastructure and applications is declared in Git, and a controller continuously reconciles the running system to match — making Git the single source of truth for deployment.

Site Reliability Engineering (SRE) is a discipline that applies software engineering practices to operations — using error budgets, SLOs, automation, and blameless incident response to balance reliability and feature velocity.

Infrastructure as Code (IaC) is the practice of provisioning and managing infrastructure using machine-readable definition files — typically in Git — rather than manual console clicks or ad-hoc scripts.

FinOps is the practice of bringing financial accountability to cloud spend — combining engineering, finance, and business stakeholders to make data-driven trade-offs between cost, speed, and quality.

Kubernetes is an open-source platform for automating deployment, scaling, and operation of containerized applications across clusters of machines — the de facto standard for container orchestration.

Zero Trust is a security model that assumes no implicit trust based on network location — every request is authenticated, authorized, and continuously validated regardless of whether it originates inside or outside the corporate perimeter.

SOC 2 is an attestation report issued by an independent auditor, verifying that a service organization’s controls meet the AICPA Trust Services Criteria — security, availability, processing integrity, confidentiality, and privacy.

Data Mesh is an organizational and architectural pattern that decentralizes ownership of analytical data to the domain teams that produce it, treating data as a product with explicit consumers, SLAs, and discoverability.

A feature store is a centralized system for managing the features used by machine learning models — providing consistent definitions, training/serving parity, freshness guarantees, and reusability across teams.

Prompt injection is an attack on LLM-powered applications where untrusted input is crafted to override the system’s intended instructions — causing the model to leak data, perform unauthorized actions, or generate harmful content.

Embeddings are numerical vector representations of text, images, audio, or other data — produced by ML models so that semantically similar items end up close together in vector space, enabling search, clustering, and retrieval.

A data contract is a formal, versioned agreement between a data producer and its consumers about the schema, semantics, freshness, and quality of a dataset — treated as code and enforced in CI rather than tribal knowledge.