AI terms, in plain English.
The vocabulary that comes up constantly in UAE business and engineering conversations about AI right now — defined clearly, without jargon, and linked to where you can actually build the skill.
AI Fundamentals
LLM (Large Language Model)
A machine learning model trained on large amounts of text to predict and generate language. LLMs (like Claude, GPT and Gemini) power most modern AI assistants and are the foundation most of HYVE's training builds on.
Related: AI Foundations for Business →Generative AI
AI that creates new content — text, code, images, audio — rather than just classifying or predicting from existing data. The umbrella term most current AI tools fall under.
Related: AI Foundations for Business →Prompt Engineering
The practice of writing instructions to an AI model to reliably get the output you want. A foundational skill for using any LLM productively, whether you're a business user or an engineer.
Related: AI Foundations for Business →Context Window
The amount of text (measured in tokens) a model can consider at once when generating a response. A larger context window lets a model work with more documents, code, or conversation history in a single pass.
Related: Software Engineer track →Multimodal AI
AI models that can process and generate more than one type of content — text, images, audio, or video — within the same system, rather than being limited to text alone.
Related: AI Foundations for Business →Hallucination
When an AI model generates confident-sounding output that is factually incorrect or unsupported by its source data. A core reason human oversight and verification remain necessary in any AI deployment.
Related: AI Governance Centre →Grounding
Techniques that connect an AI model's output to verified, real data sources — such as retrieval or tool use — to reduce hallucination and keep answers factually anchored.
Related: Software Engineer track →Inference
The process of running a trained AI model to generate a response to a given input — as distinct from training, which is how the model learned in the first place.
Related: Software Engineer track →AI Engineering
RAG (Retrieval-Augmented Generation)
An architecture that retrieves relevant information from a knowledge base or document store and feeds it to an LLM before it generates a response — used to ground answers in your organisation's actual data rather than the model's general training.
Related: Software Engineer track →Embeddings
Numerical representations of text (or other data) that capture meaning, allowing a computer to measure how similar two pieces of content are. The technical foundation RAG and semantic search are built on.
Related: Software Engineer track →Vector Database
A database optimised for storing and searching embeddings by similarity rather than exact match — the standard storage layer behind most production RAG systems.
Related: Software Engineer track →Fine-Tuning
Further training an existing AI model on a specific dataset to specialise its behaviour for a particular task or domain, rather than relying solely on prompting a general-purpose model.
Related: Software Engineer track →AI Evaluation
The systematic process of measuring how well an AI system performs against defined criteria — accuracy, relevance, safety, latency — before and after it goes into production.
Related: QA Engineer track →AI Observability
Monitoring and logging practices that let engineering teams see what an AI system is actually doing in production — inputs, outputs, tool calls, errors — so issues can be diagnosed and fixed.
Related: DevOps Engineer track →AI Agents & MCP
AI Agent
An AI system that can take multiple steps toward a goal — planning, using tools, and adapting based on results — rather than producing a single one-shot response like a chatbot.
Related: Design your first AI agent →Agentic AI
The broader shift toward AI systems that act autonomously on a user's behalf within defined boundaries, rather than only responding when prompted. The subject of the UAE's national Agentic AI programme and Dubai's private-sector mandate.
Related: Agentic AI for Business Teams →MCP (Model Context Protocol)
An open standard, introduced by Anthropic, that lets AI models connect to external tools, data sources and systems in a consistent way — instead of every AI application needing a custom integration for every tool it uses.
Related: Claude Code Training →Human-in-the-Loop
A design pattern where a human reviews, approves, or can override an AI system's actions at defined checkpoints, rather than letting it act fully autonomously. Central to responsible agent deployment.
Related: Design your first AI agent →Forward Deployed Engineer (FDE)
An engineer who embeds directly with a client to adapt and ship an AI system inside that client's real environment — combining production engineering skill with client-facing delivery. One of the fastest-growing roles in AI hiring today.
Related: FDE Readiness programme →AI Governance & Security
AI Governance
The policies, processes and oversight structures an organisation puts in place to ensure AI systems are used safely, fairly, and in line with regulation — covering everything from data handling to human oversight.
Related: AI Governance Centre →AI Safety
The field concerned with ensuring AI systems behave as intended and don't cause unintended harm, spanning both technical measures (guardrails, testing) and organisational ones (policy, review).
Related: AI Governance Centre →AI Security
Protecting AI systems and the data they access from misuse, manipulation, or exploitation — including risks specific to AI such as prompt injection and data leakage through model outputs.
Related: AI Governance Centre →