AI Glossary
A quick-reference guide to the terms shaping AI, industrial systems, and modern compute infrastructure.
Use the Glossary when you need a fast definition or refresher. If you want structured beginner learning, go to AI Basics. If you want the fastest guided path into the site, go to Start Here.
Jump to: A–C | D–I | L–O | P–Z
A–C
2nm Process
A semiconductor manufacturing node designed to increase transistor density and improve power efficiency.
Accelerated Computing
The use of specialized hardware such as GPUs and AI accelerators to run demanding workloads faster than general-purpose CPUs.
Related: GPU, Compute, HBM
Agentic Workflow
A workflow in which an AI system plans and completes multi-step tasks, often using tools, memory, and external data sources.
Related: Inference Strategy, RAG, Strategic Oversight
AI Assistant
A software assistant that uses AI to help draft, summarize, analyze, plan, or complete work tasks.
Related: Generative AI, Prompt Engineering, Workflow Automation
AI Literacy
The ability to understand, use, question, and safely apply AI tools in real work situations.
Related: Human-in-the-Loop, Strategic Oversight, Data Privacy
AI Workflow
A repeatable sequence of steps where AI supports part of the work, such as drafting, reviewing, summarizing, analyzing, or preparing a final output.
Related: Prompt Pattern, Workflow Automation, Human Review
AI-Assisted Output
Work that was created with help from AI but still shaped, reviewed, improved, and approved by a human.
Related: Human Review, Verification Loop, Responsible AI
ASIC (Application-Specific Integrated Circuit)
A custom chip built for a specific purpose, such as AI inference, networking, or encryption.
Attention Mechanism
A core transformer-model technique that helps an AI system focus on the most relevant parts of the input.
Related: Transformer, Large Language Model (LLM), Context Window
Auditable Provenance
The ability to trace where data came from, what processes were used, and how an AI-supported result was produced.
Augmentation
Using AI to expand what a person can do, rather than fully replacing the person. In careers, augmentation often means faster drafts, better analysis, stronger review, or improved decision support.
Related: Task Automation, Role Redesign, Human Review
Backpropagation
The training method neural networks use to reduce error by adjusting internal weights.
Related: Deep Learning, Neural Network, Fine-Tuning
Bandwidth
The amount of data that can move through a connection or system in a given period of time.
CAPEX
Capital expenditure. In AI, this often refers to spending on infrastructure such as chips, networking, cooling, and data centers.
Related: OPEX, ROI, Accelerated Computing
CFD / FEA
Computational Fluid Dynamics and Finite Element Analysis. These are physics-based simulation methods used to model performance, stress, heat, and flow.
Chain-of-Thought Decomposition
A structured reasoning approach that breaks a problem into intermediate steps to improve clarity and auditability.
Chiplet
A small specialized chip component that can be combined with others in a larger package.
Communication Latency
The delay between technical insight and executive understanding or action.
Compute
The processing power required to train or run AI systems.
Related: GPU, Accelerated Computing, Inference Economics
Confidential Data
Workplace, customer, financial, legal, personal, or internal information that should not be pasted into AI tools without permission and protection.
Related: Sensitive Data, Data Privacy, Responsible AI
Context Window
The amount of information an AI model can consider at one time during a prompt or interaction.
Co-Packaged Optics (CPO)
A packaging approach that places optical and electronic components close together to improve high-speed data movement.
Crosstalk
Unwanted signal interference between nearby electrical paths, which becomes more problematic at higher data rates.
D–I
Data Diode
A one-way security device that allows data to leave a network without allowing commands or traffic back in.
Data Privacy
The practice of protecting personal, company, customer, or sensitive information from improper use or exposure.
Related: Sensitive Data, Data Sovereignty, ISO 42001
Data Sovereignty
The requirement that data is governed by the laws of the country or jurisdiction where it is stored or processed.
Related: Localized RAG, ISO 42001, IEC 62443
Deep Learning
A branch of machine learning based on multi-layer neural networks.
Related: Neural Network, Backpropagation, Fine-Tuning
Design System
A set of reusable design rules, components, colors, and typographic standards used to keep digital products consistent.
Deterministic Data Flow
A communication model with predictable timing, often required in industrial and operational technology environments.
Digital Twin
A digital representation of a physical asset, process, or system that is updated with real or near-real-time data.
Related: Synthetic Twin, Predictive Maintenance, Industrial AI
Diffusion Model
A type of generative AI model commonly used to create images, video, and other media.
Edge AI
AI that runs on local devices or near the source of data instead of relying entirely on cloud processing.
Related: Industrial AI, Latency, Data Sovereignty
EUV Lithography
Extreme ultraviolet lithography used to manufacture advanced semiconductor chips.
Fab
A semiconductor fabrication facility where chips are produced.
Firmware Attestation
A cryptographic method for verifying that firmware has not been altered and is in a trusted state.
Fine-Tuning
Additional training applied to a pre-trained model to improve performance on a narrower task or domain.
Related: Foundation Model, Deep Learning, Reinforcement Learning (RL)
Foundation Model
A large pre-trained model that can be adapted for many use cases.
Front-End Scaffolding
The initial structural layout of a digital interface before refinement, testing, accessibility, and production hardening.
Generative AI
AI systems that create new content such as text, images, audio, video, or code.
Related: Large Language Model (LLM), Diffusion Model, Prompt Engineering
GPU (Graphics Processing Unit)
A processor optimized for parallel computation and widely used for AI training and inference.
Related: Accelerated Computing, HBM, Interconnect Bandwidth
Hardware Root of Trust (RoT)
A trusted hardware-based security anchor used to verify system integrity during startup and operation.
Hallucination
A confident but incorrect or unsupported AI output.
Related: RAG, Auditable Provenance, Localized RAG
HBM (High Bandwidth Memory)
A high-speed memory technology designed to feed data quickly to AI accelerators and high-performance processors.
HBM4
A newer generation of high-bandwidth memory designed for faster data throughput and more demanding AI workloads.
HPC (High-Performance Computing)
Advanced computing used to solve large, complex, or time-sensitive technical problems.
Human Review
A required check by a person before AI-supported work is sent, published, used in decisions, or shared with customers, managers, or teams.
Related: Human-in-the-Loop, Verification Loop, Responsible AI
Human-in-the-Loop
A workflow where a person reviews, guides, or approves AI-assisted work before it is used or shared.
Related: AI Literacy, Strategic Oversight, Hallucination
IEC 62443
A security framework for industrial automation and control systems.
III-V Materials
Semiconductor materials such as gallium arsenide and indium phosphide used in high-speed electronics and photonics.
Inference
The process of using a trained model to produce an output from new input.
Related: Inference Economics, Inference Strategy, Large Language Model (LLM)
Inference Economics
The cost, latency, and power profile associated with running AI models in production.
Inference Strategy
The method used to run a model, balancing factors such as speed, cost, quality, and reliability.
Industrial AI
The use of AI in manufacturing, energy, logistics, infrastructure, and other operational environments.
Related: Predictive Maintenance, Edge AI, Digital Twin
Insertion Loss
The amount of signal power lost when a component or connection is introduced into a system.
Interconnect Bandwidth
The speed at which processors, accelerators, and memory systems exchange data.
ISO 42001
An international management-system standard for governing AI responsibly.
L–O
Large Language Model (LLM)
An AI model trained on large volumes of text to understand and generate language.
Related: Transformer, Tokens, Context Window
Latency
The time delay between a request and a response.
Localized RAG
A retrieval-augmented setup in which private or local knowledge sources are used to ground AI responses without exposing sensitive data to public systems.
Related: RAG, Data Sovereignty, Hallucination
Machine Learning (ML)
A field of computing focused on systems that learn patterns from data.
Model Architecture
The structure of a model, including how layers, attention, memory, and data flow are organized.
Multimodal AI
AI that can work across more than one type of input or output, such as text, images, audio, or video.
Multimodal Portfolio
A collection of work examples that shows skill across multiple formats, such as documents, visuals, audio, video, prompts, or AI-supported workflows.
Related: Work Sample, Portfolio Signal, Skill Stack
Neural Network
A machine-learning structure inspired by connected neurons, used to detect patterns and make predictions.
NLP (Natural Language Processing)
The area of AI focused on language understanding and generation.
Nodes (Semiconductor)
Manufacturing-process generations such as 3nm or 5nm used as shorthand for chip fabrication capability.
NVLink Interconnect
A high-speed interconnect technology used to move data between GPUs and related processors.
OPEX
Operating expenditure. In AI, this often includes power, cooling, support, software, and maintenance costs.
Related: CAPEX, ROI, Inference Economics
Optical Interposer
A packaging layer that helps connect electronic and photonic components in advanced systems.
Out-of-Distribution Data
Input data that differs significantly from what a model was trained on, often reducing accuracy or reliability.
P–Z
Parameters
The internal values a model learns during training.
Pilot Chasm
The gap between a successful proof of concept and full-scale operational deployment.
Portfolio Signal
Evidence in your portfolio that helps someone quickly understand your skill, judgment, process, and ability to produce useful results with AI support.
Related: Work Sample, AI-Assisted Output, Multimodal Portfolio
Pre-Visualization (Pre-Vis)
A planning method that uses early visual mockups or 3D representations before committing to higher-cost execution.
Predictive Maintenance
Using data and AI to estimate when equipment may fail so action can be taken earlier.
Related: Industrial AI, Digital Twin, Prescriptive Guidance
Prescriptive Guidance
Recommendations generated from data or AI that suggest what action to take, not just what is likely to happen.
Probabilistic vs. Deterministic
Probabilistic systems produce likely outcomes; deterministic systems produce the same result every time for the same input.
Prompt Engineering
The practice of designing inputs that guide an AI system toward better outputs.
Related: Generative AI, Large Language Model (LLM), RAG
Prompt Pattern
A reusable way of asking AI for help, such as summarize, compare, rewrite, critique, plan, extract, or turn notes into action items.
Related: Prompt Engineering, AI Workflow, Skill Stack
RAG (Retrieval-Augmented Generation)
A method that allows an AI system to use retrieved documents or external knowledge when generating answers.
Related: Localized RAG, Hallucination, Auditable Provenance
Reinforcement Learning (RL)
A training method in which a system learns by receiving rewards or penalties based on behavior.
Responsible AI
The practice of using AI in ways that are accurate, safe, transparent, fair, reviewed, and appropriate for the workplace or audience affected by the output.
Related: Human Review, Confidential Data, Verification Loop
Role Redesign
The process of changing job responsibilities as AI takes over some tasks and creates new human responsibilities around judgment, review, coordination, and improvement.
Related: Augmentation, Task Automation, Skills Adjacency
ROI (Return on Investment)
The measurable value gained relative to the cost of an AI initiative.
Related: CAPEX, OPEX, Inference Economics
SaaS (Software as a Service)
Software delivered over the internet rather than installed and managed locally.
Scaffolding vs. Production Hardening
The difference between a quickly generated first version and a secure, tested, reliable production-grade system.
Scalability
The ability of a system to handle growing demand without unacceptable loss of performance.
Sensitive Data
Information that should be protected, such as customer records, private business data, credentials, financial details, or personal information.
Related: Data Privacy, Data Sovereignty, ISO 42001
Shadow Mode
A deployment approach in which an AI system runs alongside live operations without controlling them, allowing performance to be evaluated safely.
Skill Stack
A practical combination of skills that work together, such as communication, domain knowledge, AI prompting, analysis, workflow design, and review judgment.
Related: Prompt Pattern, AI Workflow, Multimodal Portfolio
Skills Adjacency
A nearby skill area that builds naturally from what you already know, helping you move into AI-supported work without starting from zero.
Related: Augmentation, Role Redesign, Skill Stack
Skin-Effect-Related Attenuation
Signal loss that increases at high frequencies because current travels closer to the surface of a conductor.
Strategic Oversight
Human leadership responsibility for reviewing, guiding, and governing AI-assisted work and decisions.
Synthetic Data
Artificially generated data used for training, testing, or simulation.
Synthetic Twin
A high-fidelity AI-generated representation used for visualization, communication, or scenario understanding.
Related: Digital Twin, Pre-Visualization (Pre-Vis), Industrial AI
Task Automation
Using AI or software to reduce repeated manual work inside a job. Task automation changes parts of a role before it changes the entire role.
Related: Augmentation, Workflow Automation, Role Redesign
TDP (Thermal Design Power)
A measure of the heat a processor or chip is expected to generate under normal maximum load.
Time-to-Insight (TTI)
The time from data detection to useful interpretation or action.
Tokens
The units of text an AI model processes.
Tool Stack
A small set of tools chosen to support a workflow without creating unnecessary complexity or tool overload.
Related: AI Assistant, Workflow Automation, SaaS
Transformer
The neural-network architecture behind most modern large language models.
Related: Attention Mechanism, Large Language Model (LLM), Tokens
Vera Rubin Platform
A future-facing NVIDIA architecture combining CPU and GPU advances for large-scale AI systems.
Verification Loop
A repeatable review process where AI-generated work is checked against sources, requirements, examples, facts, and human judgment before it is shared or used.
Related: Human Review, AI-Assisted Output, Responsible AI
WCAG
Web Content Accessibility Guidelines used to make web content more accessible.
Workflow Automation
Using software or AI to reduce repeated manual steps in a process so work moves faster and more consistently.
Related: AI Assistant, Prompt Engineering, Tool Stack
Work Sample
A concrete example of your ability, such as a workflow, document, prompt, analysis, before/after improvement, or project outcome that shows what you can do.
Related: Portfolio Signal, AI-Assisted Output, Multimodal Portfolio
Zero-Shot Learning
A model’s ability to handle a task it was not explicitly trained on using only prior generalized knowledge.
Where to Go Next
Use one of these paths when you are ready to move from definitions into guided learning and current coverage.