The MetaSys AI Glossary
Plain-language definitions of the AI, automation, data, and cloud terms we use across our work. Written for decision-makers, not just engineers, and linked to the capabilities and guides where each idea shows up in practice.
Agentic AI & LLMs
Agentic AI
AI systems that plan, decide, and take action across multi-step workflows with limited human input, instead of only answering a single prompt. They monitor, evaluate options, and act inside your tools.
Agentic AI SystemsLarge Language Model (LLM)
A model trained on large volumes of text that predicts and generates language. LLMs power reasoning, summarization, classification, and the planning loop inside agents.
Multi-Agent System
An architecture where specialist agents work in sequence or in parallel, handing off tasks with shared memory and state. Used when one agent cannot own an end-to-end process alone.
Multi-Agent Systems guideRetrieval-Augmented Generation (RAG)
A pattern that retrieves relevant documents from your own data and feeds them to the model at query time, so answers are grounded in your knowledge rather than the model's training alone.
The RAG retrieval problemHuman-in-the-Loop
A design where the agent pauses at defined gates for a person to review, approve, or correct an action before it executes. It keeps autonomy bounded where the cost of error is high.
Fine-Tuning
Further training of a base model on domain-specific examples so it performs better on your task, terminology, and edge cases than a general-purpose model.
Context Window
The amount of text a model can consider at once. It limits how much history, retrieved data, and instruction the model can use in a single step, and it shapes how systems are architected.
Automation
Intelligent Automation
Automation that combines rules with machine learning so it can handle variation, unstructured inputs, and judgement, not just fixed scripts.
AI & Intelligent AutomationRobotic Process Automation (RPA)
Rule-based bots that mimic clicks and keystrokes to move data between systems. Reliable for stable, structured tasks, but brittle when inputs change.
RPA vs AI automationWorkflow Orchestration
Coordinating multiple steps, tools, and approvals into one reliable flow with error handling, retries, and observability.
AI workflow automation guideIntelligent Process Automation (IPA)
An umbrella for combining RPA, AI, and orchestration to automate end-to-end business processes rather than isolated tasks.
IPA guideData & Machine Learning
Data Platform
The ingestion, storage, transformation, and serving layer that makes data usable for analytics and AI. Without it, AI projects stall on data access and quality.
Data & AI PlatformsData Lakehouse
An architecture that combines the low-cost storage of a data lake with the structure and performance of a warehouse, so one platform serves both analytics and machine learning.
Data lakehouse architectureVector Database
A store optimized for embeddings, the numeric representations of text and images. It powers semantic search and the retrieval step in RAG systems.
MLOps
The practices and tooling for deploying, monitoring, and retraining machine learning models in production, the way DevOps does for software.
Real-Time Pipeline
A streaming data flow that processes events as they happen, enabling live dashboards, monitoring agents, and instant decisions instead of overnight batches.
Real-time pipeline guideCloud & Delivery
Cloud-Native
Software built to run on cloud infrastructure using containers, managed services, and automation, so it scales elastically and ships frequently.
Cloud & DevOps EngineeringDevOps
A practice that unifies development and operations with automated build, test, and deploy pipelines to release software faster and more safely.
DevOps for AI teamsGlobal Capability Center (GCC)
A dedicated offshore engineering team that operates as an extension of your company, with your processes and standards, rather than a per-project vendor.
Global Capability CentersObservability
The ability to understand a system's internal state from its outputs, traces, logs, and metrics. For AI systems it also covers accuracy, latency, cost, and drift.
Turn these ideas into working systems.
We design, build, and operate the systems behind these terms. Bring a problem and we will map a concrete path to production.