AI Vocab
Understand every AI term, in context
Most glossaries define a word and stop. Every term here sits on the same map as the rest, so it also tells you what to learn first, what it unlocks, and which careers run through it.
88 terms, 34 with a full explanation written so far. The rest carry their meaning, their place on the map and their sources, and say plainly that the longer write-up is still to come.
88 terms
Activation Functions
Explanation comingThe non-linearity without which a deep network collapses into a linear one.
Agent Evaluation
Explanation comingJudging a trajectory, not an answer. Compounding error is the thing to measure.
Agent Memory
Explanation comingShort-term scratchpads, long-term stores, and deciding what is worth remembering.
AI Agents
Models that decide what to do next, call tools, and act in a loop rather than answering once.
AI Architecture
System-level design: model routing, boundaries, failure modes, governance. The summit node.
AI Engineering
The bridge from a working notebook to a system real users depend on.
Analytics Engineering
Explanation comingSoftware practice applied to transformation: version control, tests and review on the models analysts use.
Artificial Intelligence
The root of the Atlas. Everything else in this map is a descendant of this one idea.
Attention
Letting every position look at every other position. The hinge the modern field turns on.
Backpropagation
The chain rule applied at scale. How a network learns which weights were at fault.
Batch and Streaming
Bounded versus unbounded data. Latency you can promise against complexity you have to operate.
Calculus & Optimisation
Explanation comingDerivatives, the chain rule, and what it means to walk downhill in a loss landscape.
Change Data Capture
Explanation comingReading the database log instead of polling the table. How a warehouse stays minutes behind production.
Chunking
How you split documents. Where most RAG systems quietly lose their quality.
Cloud Data Platforms
Explanation comingStorage, compute, IAM and the bill. Pick one cloud properly rather than three of them badly.
Columnar Storage Formats
Explanation comingParquet, Iceberg, Delta. Row groups, predicate pushdown, and why layout decides your query bill.
Computer Vision
Getting structured meaning out of pixels: classification, detection, segmentation.
Context Windows
Explanation comingThe model’s entire working memory, and the constraint that shapes most system design.
Convolutional Networks
Explanation comingWeight sharing over spatial structure. Still the efficient default for images.
Cost & Latency
Explanation comingToken economics, model routing and caching. The constraint that decides architecture.
Cross-Validation
Explanation comingGetting an honest performance estimate when you do not have data to spare.
Data Engineering
Designing the systems that make data trustworthy, timely and cheap enough for analytics and AI to run on.
Data Infrastructure and CI/CD
Explanation comingContainers, IaC and a deploy that runs your transformations against a staging warehouse before production.
Data Modelling
Choosing the grain. The decision that determines every query anyone will ever write against your tables.
Data Pipelines
Moving data from where it is produced to where it is asked about, reliably and on a schedule.
Data Quality
Explanation comingTests on data, not just on code. The pipeline that succeeds every night while writing zero rows.
Data Strategy
Explanation comingProvenance, licensing, freshness and permissions. The unglamorous moat.
Data Warehouse
Columnar storage and separated compute. Why a scan of a billion rows can cost cents or hundreds of dollars.
Data Wrangling
Explanation comingLoading, cleaning and reshaping data. Realistically most of the job.
Decoding Strategies
Explanation comingTemperature, top-p, beam search. How a distribution becomes the text you see.
Deep Learning
Many-layered neural networks that learn their own features instead of being handed them.
Diffusion Models
Explanation comingLearn to reverse noise, then run it backwards from pure noise to generate images.
Distributed Processing
Explanation comingSpark and its relatives. Partitions, shuffles, and the skew that leaves one task running for six hours.
Embeddings
Meaning as coordinates. The representation that makes search, clustering and RAG possible.
ETL and ELT
The transform step moved. That single change is what the warehouse era is actually about.
Feature Engineering
Explanation comingShaping raw data into signal. Where domain knowledge earns its keep.
Fine-Tuning
Adapting a pretrained model to your task. Often the wrong first answer.
Generative AI
Models that produce new artefacts rather than predicting a label. The densest region of the Atlas.
Git & Collaboration
Explanation comingVersion control, review and reproducibility. Assumed silently by every team.
Governance and Lineage
Explanation comingWho may read this, where it came from, and what breaks if it changes. The questions auditors and outages both ask.
Gradient Descent
The optimisation procedure underneath essentially every model in this Atlas.
Guardrails & Safety
Explanation comingInput validation, output filtering, prompt injection, and where to put the boundary.
Hybrid Search
Explanation comingCombining keyword and semantic retrieval, because each fails where the other works.
Image Segmentation
Explanation comingPer-pixel classification, for when a bounding box is not precise enough.
Inference Optimisation
Explanation comingQuantisation, caching, speculative decoding. Buying back latency and margin.
Lake and Lakehouse
Table formats put warehouse guarantees on object storage. What the lakehouse actually adds over a folder of files.
Large Language Models
Transformers trained on enough text to become general-purpose reasoning surfaces.
Linear & Logistic Regression
Explanation comingThe two models you should always try first, and often ship.
Linear Algebra
Explanation comingVectors, matrices and the operations every model is secretly made of.
LLM Evaluation
Measuring quality when there is no single correct output. The hardest unsolved problem in shipping.
LoRA & PEFT
Explanation comingTrain a few million parameters instead of a few billion, and lose almost nothing.
Machine Learning
Systems whose behaviour is learned from data rather than written as rules.
Math for AI
The specific mathematics that appears in practice, and nothing beyond it.
Model Context Protocol
Explanation comingA standard interface between models and the tools and data they need.
Model Evaluation
The skill that separates people who ship models from people who publish notebooks.
Model Serving
Explanation comingAPIs, batching, streaming, and the deployment surface between model and product.
Multi-Agent Systems
Explanation comingSeveral specialised agents coordinating. Powerful, and usually premature.
Multimodal AI
Explanation comingOne model, several senses. Text, image, audio and video in a shared representation.
Neural Networks
Layers, weights and activations. The unit of construction for everything downstream.
Object Detection
Explanation comingWhat is in the image, and where. Boxes, anchors and non-maximum suppression.
Observability & Evals
Tracing, logging and regression suites. Knowing you broke it before a user tells you.
Optimisers
Explanation comingAdam, AdamW, schedules and warmup. What you actually tune when training stalls.
Orchestration
Explanation comingDependencies, retries and backfills. Where a folder of scripts becomes a system you can reason about.
Overfitting & Regularisation
Explanation comingWhy a model that memorises the training set is worse than one that does not.
Pipeline Observability
Explanation comingFreshness, volume and cost as monitored signals, so a silent failure is loud before a stakeholder finds it.
Planning & Reasoning
Explanation comingDecomposition, reflection, and letting the model check its own work.
Pretraining
Explanation comingThe expensive part: corpus, curriculum, scaling laws, and the compute bill.
Probability & Statistics
Explanation comingDistributions, expectation, and why every evaluation number needs an error bar.
Production RAG
Ingestion pipelines, freshness, permissions, caching and cost at real volume.
Prompt Engineering
Specifying a task precisely enough that a probabilistic system does it reliably.
Python
The only genuinely non-negotiable prerequisite in the entire Atlas.
RAG
Give the model the right documents at query time instead of hoping it memorised them.
RAG Evaluation
Explanation comingSeparating retrieval failures from generation failures, so you fix the right one.
Recurrent Networks
Explanation comingRNNs and LSTMs. Largely superseded, but the reason attention was invented.
Reranking
Explanation comingA second, slower pass over the top results. Usually the largest single quality win available.
RLHF & Preference Tuning
Explanation comingAligning output with human judgement when there is no correct answer to train on.
SQL
Explanation comingWhere the training data actually lives, and the fastest way to interrogate it.
Supervised Learning
Learning a mapping from labelled examples. The workhorse of applied ML.
Tokenisation
Explanation comingHow text becomes numbers, and why the model cannot count the letters in a word.
Tool Use
Function calling, schemas, and validating what the model asks you to run.
Training at Scale
Explanation comingData, tensor and pipeline parallelism, mixed precision, and the memory wall.
Transfer Learning
Explanation comingStart from someone else’s trained weights. The economics of modern AI in one idea.
Transformers
Attention, feed-forward layers, residuals and normalisation. The block that ate the field.
Trees & Ensembles
Explanation comingGradient boosting still wins on tabular data. This is not a historical note.
Unsupervised Learning
Explanation comingFinding structure in data nobody labelled: clustering, density, dimensionality.
Vector Databases
Explanation comingWhere embeddings live: indexes, filters, and the operational reality of them.
Vector Search
Explanation comingApproximate nearest neighbours at scale. HNSW, IVF, and the recall/latency trade.
Vision-Language Models
Explanation comingModels that read an image and talk about it. Where vision rejoined the mainstream.
Not sure which of these matter for you?
The free assessment places you on the same map and names the terms standing between you and the role you want.
Take the free assessment