- Accuracy, precision and recall Machine learning
- Accuracy: share of correct predictions. Precision: of the items flagged, how many were right. Recall: of the real items, how many were found.
- Agent memory Building with AI
- Information an agent keeps between steps or sessions, such as user preferences or past results.
- API Deep learning & LLMs
- A way for programs to talk to each other. Model APIs let your code send prompts and receive replies.
- Artificial intelligence (AI) Foundations
- Computer systems that perform tasks that normally need human intelligence, such as understanding language, recognising images or making decisions.
- Bias AI security
- Systematic unfairness in a model's outputs, often learned from unrepresentative data.
- Chain of thought / reasoning Deep learning & LLMs
- Getting a model to work through intermediate steps before answering. "Thinking" models do this internally.
- Chatbot Foundations
- An application that lets people talk to an AI model in natural language. The model is the engine; the chatbot is the product around it.
- Classification Machine learning
- Predicting a category, such as spam or not spam.
- Confusion matrix Machine learning
- A table of true positives, false positives, true negatives and false negatives for a classifier.
- Context window Deep learning & LLMs
- The maximum number of tokens a model can consider at once, including your prompt and its reply.
- Convolutional neural network (CNN) Machine learning
- A network designed for images that learns filters to detect edges, shapes and objects.
- Dataset Python & data
- A collection of examples used to train or test a model, such as labelled images or text documents.
- Deep learning Foundations
- Machine learning with neural networks that have many layers. It powers modern image, speech and language AI.
- Diffusion model Deep learning & LLMs
- The kind of model behind most image generators: it learns to turn random noise into an image step by step.
- Epoch Machine learning
- One full pass through the training data.
- Feature Python & data
- A measurable input property used by a model, such as house size or number of bedrooms.
- Few-shot prompting Deep learning & LLMs
- Including a few worked examples in the prompt so the model copies the pattern.
- Fine-tuning Deep learning & LLMs
- Further training a pretrained model on a smaller, specific dataset to change its behaviour or specialise it.
- Generative AI Foundations
- AI that creates new content — text, images, audio, video or code — rather than only classifying or predicting.
- Hallucination Foundations
- When a model states something false or made-up with confidence. It happens because models generate plausible text, not verified facts.
- Human in the loop Building with AI
- Requiring a person to review or approve an AI's action before it takes effect.
- Inference Foundations
- Using a trained model to make predictions or generate output. Every chatbot reply is an inference.
- Label Python & data
- The correct answer attached to a training example, such as "spam" or "not spam".
- Machine learning (ML) Foundations
- A way of building AI where the system learns patterns from examples (data) instead of following hand-written rules.
- Matrix Python & data
- A grid of numbers. Neural network layers are mostly matrix multiplications.
- Model Context Protocol (MCP) Building with AI
- An open standard for connecting AI applications to tools and data sources through MCP servers. MCP tool scanner →
- Multi-agent system Building with AI
- Several agents with different roles that work together on a task.
- Multimodal model Deep learning & LLMs
- A model that handles more than one type of input or output, such as text and images.
- Notebook (Jupyter / Colab) Python & data
- An interactive document that mixes code, results and notes. The standard way to experiment in ML.
- Orchestration Building with AI
- The code that runs an agent loop: calling the model, executing tools, managing memory and stopping conditions.
- Pretraining Deep learning & LLMs
- The first, most expensive training stage, where a model learns to predict the next token on a huge text corpus.
- Prompt Foundations
- The input you give a model: instructions, questions, examples and context.
- Prompt engineering Deep learning & LLMs
- Designing prompts — clear instructions, examples, structure — to get reliable results from a model.
- Regression Machine learning
- Predicting a number, such as tomorrow's electricity demand.
- Reinforcement learning Machine learning
- Learning by trial and error, receiving rewards or penalties for actions.
- Responsible AI AI security
- Building AI that is fair, transparent, safe, private and accountable.
- RLHF Deep learning & LLMs
- Reinforcement learning from human feedback: training a model to prefer the answers people rate as better.
- Supervised learning Machine learning
- Learning from labelled examples, for example predicting price from house features.
- Train / validation / test split Machine learning
- Dividing data so you train on one part, tune on another and measure final performance on data the model never saw.
- Training Foundations
- The process of adjusting a model's parameters so its predictions match the examples in the training data.
- Unsupervised learning Machine learning
- Finding structure in data without labels, for example grouping customers into segments.
- Vector database Deep learning & LLMs
- A database that stores embeddings and finds the most similar ones quickly. Used for RAG.
- Weights and biases Machine learning
- The numbers inside a neural network that training adjusts. Together they are the model's parameters.