Home

AI & Data Terms

Plain-language explanations of data and AI terminology.

96 terms

Activation Function

A nonlinear function applied to neuron outputs — ReLU, sigmoid, GELU — that lets networks learn complex patterns.

Deep Learning · 1 min read

AI Agent

An LLM-powered system that plans steps, uses tools, and acts toward a goal with minimal human intervention.

AI Agents · 2 min read

Artificial Intelligence (AI)

Systems that perform tasks that typically require human intelligence — reasoning, perception, language, and decision-making.

AI Foundations · 1 min read

Attention Mechanism

A way for models to focus on relevant parts of the input when producing each output — the core idea behind transformers.

Deep Learning · 1 min read

Autoencoder

A neural network trained to compress input into a latent code and reconstruct it — useful for representation learning and anomaly detection.

Deep Learning · 1 min read

Backpropagation

The algorithm that computes how each neural network weight contributed to the error so gradient descent can update them.

Deep Learning · 1 min read

Batch Normalization

A technique that normalizes layer inputs across a mini-batch to stabilize and often speed up training.

Deep Learning · 1 min read

Bias in AI

Systematic unfairness or skewed behavior in data and models that harms some groups or languages more than others.

AI Foundations · 1 min read

Bias–Variance Tradeoff

The tension between models that are too rigid (high bias) and models that are too sensitive to training noise (high variance).

Machine Learning · 1 min read

Big Data

Datasets so large or fast that they need distributed storage and processing beyond a single machine's comfort zone.

Data Science · 1 min read

Chain of Thought

Prompting or training models to reason step by step before answering — often improving multi-step accuracy.

Prompt Engineering · 1 min read

CI/CD

Continuous Integration and Continuous Delivery — automating build, test, and deployment every time code changes.

DevOps · 1 min read

Classification

Predicting a discrete category or label for an input — spam vs not spam, language ID, sentiment class.

Machine Learning · 1 min read

Cloud Computing

Renting compute, storage, and managed services over the internet instead of owning physical servers.

Cloud Computing · 1 min read

Clustering

Grouping similar data points together without predefined labels — a core unsupervised technique.

Machine Learning · 1 min read

Computer Vision

The field of AI that enables machines to interpret images and video — detect, classify, segment, and describe visual content.

AI Foundations · 1 min read

Context Window

The maximum amount of text (in tokens) an LLM can consider in a single request — prompt plus response.

Large Language Models · 2 min read

Convolutional Neural Network (CNN)

A neural architecture that slides filters over grid-like data — images, spectrograms — to detect local patterns.

Deep Learning · 1 min read

Cross-Validation

A method to estimate how well a model generalizes by training and testing on multiple splits of the data.

Machine Learning · 1 min read

Data Labeling

Annotating examples with the correct tags, transcripts, boxes, or answers so supervised models can learn.

Data Science · 1 min read

Data Lake

A central store for large amounts of raw data in native formats — files, logs, images — before heavy structuring.

Data Science · 1 min read

Data Pipeline

An automated flow that moves and transforms data from sources to storage, analytics, or model training.

Data Science · 1 min read

Data Quality

How accurate, complete, consistent, and timely your data is — the hidden limiter of every ML system.

Data Science · 1 min read

Data Warehouse

A structured analytical database optimized for querying cleaned, modeled business and product data.

Databases · 1 min read

Dataset

A structured collection of examples used to train, validate, or evaluate machine learning models.

Data Science · 1 min read

Deep Learning

Machine learning using neural networks with many layers that learn hierarchical representations from data.

Deep Learning · 1 min read

Diffusion Model

A generative model that learns to reverse a gradual noising process — the engine behind many modern image and audio generators.

Deep Learning · 1 min read

DNS

Domain Name System — translates human-readable domain names like goobolabs.so into IP addresses computers use.

Networking · 1 min read

Docker

A platform for packaging applications and dependencies into portable containers that run consistently anywhere.

DevOps · 2 min read

Dropout

A regularization technique that randomly disables neurons during training to reduce overfitting.

Deep Learning · 1 min read

Embedding Model

A model specialized in mapping text (or other inputs) to vectors optimized for similarity, retrieval, and clustering.

Large Language Models · 1 min read

Embeddings

Dense numerical vectors that represent text, images, or other data so similar items sit close together in vector space.

AI Foundations · 2 min read

Encryption

Encoding data so only parties with the correct key can read it — essential for privacy and security.

Cybersecurity · 1 min read

Ensemble Learning

Combining multiple models to get better predictions than any single model alone — bagging, boosting, stacking.

Machine Learning · 1 min read

ETL / ELT

Extract, Transform, Load (or Extract, Load, Transform) — patterns for moving data from sources into analytics or ML systems.

Data Science · 1 min read

Exploratory Data Analysis (EDA)

Investigating a dataset with summaries and plots to understand distributions, issues, and promising signals before modeling.

Data Science · 1 min read

Feature Engineering

Creating informative input variables from raw data so models can learn patterns more effectively.

Data Science · 1 min read

Feature Store

A system for defining, storing, and serving ML features consistently offline for training and online for inference.

Data Science · 1 min read

Few-shot Learning

Getting a model to perform a task from just a handful of examples — often provided in the prompt for LLMs.

Large Language Models · 1 min read

Fine-tuning

Adapting a pre-trained model to a specific task or domain by training further on a smaller, targeted dataset.

Machine Learning · 2 min read

Foundation Model

A large model trained on broad data that can be adapted to many downstream tasks — the base layer of modern AI systems.

AI Foundations · 1 min read

Function Calling

Letting an LLM request structured tool calls — APIs, calculators, databases — instead of only generating free text.

AI Agents · 1 min read

GAN

Generative Adversarial Network — a generator and discriminator compete so the generator learns to create realistic synthetic data.

Deep Learning · 1 min read

Generative AI

AI systems that create new content — text, images, audio, code — rather than only classifying or scoring inputs.

AI Foundations · 1 min read

Git

A version control system that tracks code changes, enables collaboration, and supports branching and rollback.

Programming · 1 min read

GPU

Graphics Processing Unit — parallel hardware that accelerates training and inference for neural networks.

AI Foundations · 1 min read

Gradient Descent

The workhorse optimization algorithm that updates model weights by following the slope of the loss downhill.

Machine Learning · 1 min read

GraphQL

A query language for APIs that lets clients request exactly the fields they need in a single request.

Backend Development · 1 min read

Grounding

Connecting model outputs to verifiable sources — documents, tools, or databases — so answers stay tethered to evidence.

Large Language Models · 1 min read

Hallucination

When an LLM generates confident, plausible-sounding text that is factually wrong or unsupported.

Large Language Models · 1 min read

HTTP

Hypertext Transfer Protocol — the foundation of how browsers and apps request and send data on the web.

Networking · 1 min read

Hyperparameter

Settings you choose before training — learning rate, batch size, layers — that are not learned from the data itself.

Machine Learning · 1 min read

Inference

Running a trained model on new data to produce predictions or generated output — the "using" phase of ML.

Machine Learning · 1 min read

JavaScript

The programming language of the web — runs in browsers and on servers (Node.js) for full-stack development.

Programming · 1 min read

Knowledge Distillation

Training a smaller student model to mimic a larger teacher — compressing capability into a cheaper model.

Machine Learning · 1 min read

Kubernetes

An orchestration platform that runs, scales, and heals containerized applications across clusters of machines.

DevOps · 1 min read

Large Language Model (LLM)

A Transformer trained on massive text corpora to predict and generate language — the engine behind ChatGPT-style assistants.

Large Language Models · 2 min read

LoRA

Low-Rank Adaptation — a parameter-efficient fine-tuning method that trains small adapter matrices instead of all model weights.

Large Language Models · 1 min read

Loss Function

A score that measures how wrong a model's predictions are — training tries to make this number smaller.

Machine Learning · 1 min read

LSTM

Long Short-Term Memory — a recurrent architecture with gates that better remember long-range information than vanilla RNNs.

Deep Learning · 1 min read

Machine Learning

Systems that improve performance on a task by learning patterns from examples instead of explicit rules.

Machine Learning · 2 min read

Microservices

An architecture where an application is split into small, independently deployable services that communicate over the network.

Backend Development · 1 min read

MLOps

Practices and tools for deploying, monitoring, and iterating on ML models reliably in production — DevOps for machine learning.

Machine Learning · 1 min read

Model Evaluation

Measuring how well a model performs — metrics, benchmarks, human preference tests, and real-world online evaluation.

Machine Learning · 1 min read

Multimodal AI

Models that understand or generate across multiple modalities — text, images, audio, video — in one system.

AI Foundations · 1 min read

Natural Language Processing (NLP)

The field of AI focused on understanding, generating, and transforming human language with computers.

AI Foundations · 1 min read

Neural Network

A computing system of layered nodes that learns patterns by adjusting connection strengths through training.

Deep Learning · 1 min read

Normalization & Standardization

Rescaling numeric features so different units and ranges do not dominate training — min-max, z-score, and related methods.

Data Science · 1 min read

OAuth

An authorization standard that lets apps access user data on another service without sharing passwords.

Cybersecurity · 1 min read

Overfitting

When a model memorizes training data noise instead of learning patterns that generalize to new examples.

Machine Learning · 1 min read

Parquet

A columnar file format popular for analytics and ML datasets — efficient compression and fast column reads.

Data Science · 1 min read

Prompt Engineering

The practice of crafting instructions, examples, and context so LLMs produce reliable, useful outputs.

Prompt Engineering · 2 min read

Python

A readable, versatile programming language used everywhere from data science to web backends and AI tooling.

Programming · 2 min read

Quantization

Storing and computing model weights with fewer bits (e.g., 8-bit or 4-bit) to shrink memory use and often speed inference.

Large Language Models · 1 min read

Random Forest

An ensemble of decision trees trained on random subsets of data and features — robust and widely used on tabular data.

Machine Learning · 1 min read

Recurrent Neural Network (RNN)

A neural architecture that processes sequences step by step, carrying a hidden state through time — precursor to LSTMs and transformers.

Deep Learning · 1 min read

Regression

Predicting a continuous numeric value — price, temperature, score, or duration — from input features.

Machine Learning · 1 min read

Reinforcement Learning (RL)

Learning by trial and error — an agent takes actions, receives rewards, and improves a policy over time.

Machine Learning · 1 min read

REST API

A style of web API that uses HTTP methods and URLs to create, read, update, and delete resources.

Backend Development · 2 min read

Retrieval-Augmented Generation (RAG)

An architecture that retrieves relevant documents first, then asks an LLM to answer using that grounded context.

Large Language Models · 2 min read

RLHF

Reinforcement Learning from Human Feedback — aligning models with human preferences using ranked examples and a reward model.

Large Language Models · 1 min read

Semantic Search

Search that matches meaning and intent, not just exact keywords — powered by embeddings and similarity.

AI Foundations · 2 min read

Serverless

A cloud model where you run code in response to events without managing servers — billing per execution.

Cloud Computing · 1 min read

SQL

Structured Query Language — the standard way to store, query, and manage relational data in databases.

Databases · 1 min read

Streaming Data

Continuously generated events processed in near real time — logs, clicks, sensor readings — rather than only in daily batches.

Data Science · 1 min read

Supervised Learning

Training a model on labeled examples where each input is paired with the correct output.

Machine Learning · 1 min read

System Prompt

Instructions that set an LLM's role, rules, and behavior for a session — usually higher priority than user messages.

Prompt Engineering · 1 min read

Temperature

A sampling parameter that controls how random or focused an LLM's next-token choices are.

Large Language Models · 1 min read

Tokenization

Splitting text into subword units (tokens) that language models read, process, and generate.

Large Language Models · 2 min read

Train / Validation / Test Splits

Separating data so you train on one set, tune on another, and report honest performance on a final untouched test set.

Machine Learning · 1 min read

Transfer Learning

Reusing knowledge from a model trained on one task or dataset to improve performance on a related task with less data.

Deep Learning · 1 min read

Transformers

A neural network architecture that uses attention to model relationships between all parts of a sequence at once.

Deep Learning · 2 min read

Underfitting

When a model is too simple to capture the real patterns in the data — poor performance on both train and test.

Machine Learning · 1 min read

Unsupervised Learning

Finding structure in data without labeled answers — clustering, dimensionality reduction, and pattern discovery.

Machine Learning · 1 min read

Vector Database

A database optimized for storing embedding vectors and retrieving nearest neighbors by similarity at scale.

Databases · 2 min read

Zero-shot Learning

Performing a task with no task-specific examples — only instructions or class descriptions.

Large Language Models · 1 min read

Frequently asked

What is this glossary?

Plain-language explanations of AI and data terms — from tokenization to RAG — written for people without a computer science background.

Are the terms ranked by difficulty?

Yes — each entry is tagged Beginner, Intermediate, or Advanced, with related terms linked so you can follow a topic from fundamentals to depth.

Is it available in Somali?

Yes — the glossary itself is translated, not just the site navigation. Switch languages from the header to read it in Somali.