Data and AI glossary
Words from Maestro's courses on data and AI, each explained in a sentence or two.
/
AI agents25 words
- Agent harnessThe program that wraps a language model and makes it an agent: it gives the model tools, asks for your permission, keeps the session and decides what goes into its context.
- Agent instructions fileA text file in your project that the agent's harness loads at the start of every session, so the agent knows from the start what it always needs to know: rules, facts about the data and how to check its work.
- Agentic AI, has a figureAgentic AI works toward a goal on its own over several steps: it plans, uses tools like files, search or code, checks the result and decides what to do next, instead of answering once.
- AI agentA program that pairs a language model with tools, such as terminal commands and files, to carry out a task in several steps: it reads its context, acts, looks at the result and repeats.
- AI hallucination, has a figureA hallucination is an answer from an AI model that sounds confident and well written but is false or comes from no source at all: a number, a fact or a quote nobody ever said.
- Claude Code, has a figureClaude Code is Anthropic's AI agent that works on your computer: it reads your project's files, runs commands and makes changes over several steps, asking your permission for what matters. You use it from the terminal, the desktop app or editors like VS Code.
- Claude Cowork, has a figureClaude Cowork is the mode of the Claude desktop app where Claude works with the files in a folder you give it: it reads, organizes and creates documents and spreadsheets over several steps, with no coding or terminal required.
- Claude for Excel, has a figureClaude for Excel is Anthropic's add-in that puts Claude inside Excel: it reads your workbook, answers with citations to the cells, writes formulas and makes changes in the same file.
- Claude skills (Agent Skills), has a figureClaude skills (Agent Skills) are folders of instructions, each with a SKILL.md file, that teach Claude or another AI agent a procedure so it does the task your way.
- Context engineering, has a figureContext engineering is choosing what goes into an AI model's context window at each step (instructions, files, tool results, history) so it has what it needs and little else.
- Context windowEverything a language model has in front of it at a given moment: your instruction, the files it has read and the results of its tools. What isn't in the context, the model can't use.
- Copilot in Excel, has a figureCopilot in Excel is Microsoft's AI assistant inside Excel: you ask for an analysis in your own words, and it builds formulas, PivotTables, charts or summaries in your workbook.
- Fine-tuning, has a figureFine-tuning is training an existing AI model a little more on your own examples, so it gets better at one kind of task, without building a model from scratch.
- Generative AI, has a figureGenerative AI is artificial intelligence that creates new content, like text, images, code or audio, from patterns it learned in huge numbers of examples. ChatGPT and Claude are examples.
- LLM (large language model), has a figureAn LLM is a program trained on huge amounts of text to continue a piece of text one chunk at a time: it works out which chunk is likeliest to come next and writes it. ChatGPT, Claude and Gemini all run on an LLM.
- Machine learning, has a figureMachine learning has a program learn rules from examples that come with answers, instead of someone writing them, so it can predict new cases.
- MCP (Model Context Protocol)An open standard for connecting AI agents to outside tools and data sources, such as databases, APIs or documents, through one kind of connector.
- n8n, has a figuren8n is a workflow automation tool: you connect apps, code and AI steps as nodes on a canvas, and a trigger, like a schedule or a new email, runs the workflow on its own.
- Neural network, has a figureA neural network is a model made of layers of simple math units that learns patterns from examples by adjusting many numbers, called weights. It's what LLMs are built on.
- Prompt, has a figureA prompt is what you write to an AI model to ask it for something: the question or instruction, plus everything you give it to work with, like data, examples or the format you want back.
- Prompt engineering, has a figurePrompt engineering is writing, testing and revising the instructions for a language model until it does a task reliably, measured against cases with known answers.
- RAG (retrieval-augmented generation), has a figureRAG (retrieval-augmented generation) gets an AI to answer from your documents: it first searches them for the passages closest to the question and hands those to the model with the prompt.
- Token (in AI), has a figureA token is the piece of text a language model reads and writes in: sometimes a whole word, sometimes part of a word, a number or a punctuation mark. Models count their limits and their cost in tokens.
- Vector database, has a figureA vector database stores embeddings, lists of numbers that capture what a text or image means, and finds the ones closest to a query's, so you search by meaning instead of exact words.
- Vibe coding, has a figureVibe coding is building software by describing what you want to an AI and accepting the code it writes without reading it, judging only by whether the result seems to work.
Charts20 words
- AltairAltair (Vega-Altair) is a Python package for making charts by describing them: you say which column goes on each axis, color or size, and it writes the Vega-Lite that draws the chart.
- Bar chart, has a figureA bar chart compares amounts across categories: each bar is a category and its length is the number. Use it to see which branch sold the most.
- Box plot, has a figureA box plot sums up a variable in five numbers: the minimum, the first quartile, the median, the third quartile and the maximum. The box runs from the first to the third quartile and holds the middle half of the data.
- Conditional formatting, has a figure, says how to do itConditional formatting changes a cell's color or style based on its value, with rules like "greater than", "duplicates" or a color scale, and updates on its own when the data changes.
- Correlation, has a figureCorrelation measures how closely two numbers move together, from −1 to 1. Near 1: when one rises, so does the other. Near 0: no straight-line link.
- Correlation vs. causation, has a figureCorrelation means two numbers move together; causation means changing one changes the other. Moving together doesn't prove that one causes the other.
- Dashboard, has a figureA dashboard puts the few numbers someone checks often to make a decision on one screen, each next to its goal or its trend.
- Data visualization, has a figureData visualization is showing data as pictures, like charts, maps and heat maps, so the eye can see patterns, comparisons and outliers that a table hides.
- Gantt chart, has a figureA Gantt chart puts a project's tasks in rows, each one a bar on a calendar: when it starts, how long it takes and what it waits for.
- Heat map, has a figureA heat map is a grid of colored cells where the shade shows the size of each value, like sales by day and hour, so you can see at a glance where activity bunches up.
- Histogram, has a figureA histogram shows how the values in a column of numbers are spread out: it cuts them into ranges of equal width and draws one bar per range, as tall as the number of values that fall in it.
- Line chart, has a figureA line chart joins points in order, almost always over time, to show how a number changes: whether it rises, falls or repeats every season.
- Linear regression, has a figure, says how to do itLinear regression draws the straight line that best follows the points of two variables, to estimate one from the other: y = a + b·x.
- Log scaleA log scale is an axis where each equal step multiplies the value instead of adding to it, so 1,000, 10,000 and 100,000 sit evenly spaced.
- Pareto chart, has a figureA Pareto chart sorts a problem's causes from largest to smallest, with a line of their running total, to show the few causes behind most of it.
- Pie chart, has a figureA pie chart shows how a total splits up: each slice is one part. It works with a few parts that add up to 100%.
- Power BI, has a figurePower BI is Microsoft's tool for connecting data from several sources and building interactive reports and dashboards your team opens online.
- Scatter plot, has a figureA scatter plot draws one dot per case, with one variable on each axis, to show whether two numbers move together: when one goes up, does the other go up, go down, or do nothing?
- Time series, has a figureA time series is a list of measurements of the same thing taken in order, every day, month or year, like daily sales or monthly inflation.
- Vega-LiteVega-Lite is a language for describing charts in JSON: you name the data, the mark and which column goes on each axis, and a renderer draws it. Altair writes Vega-Lite for you.
Checking your data31 words
- Big data, has a figureBig data is data so large, fast or varied that it doesn't fit on one computer, so it's stored and processed across many machines.
- COUNTIF, has a figure, says how to do itCOUNTIF is an Excel and Google Sheets function that counts the cells in a range that meet one condition: =COUNTIF(range, criteria).
- Data analytics, has a figureData analytics is the work of turning data into answers: asking a question, cleaning the data, running the numbers, and explaining what happened, why, and what to do next.
- Data cleaning, has a figureData cleaning is finding and fixing what's wrong in a table before you analyze it: duplicates, blanks, typos and formats that don't match.
- Data governance, has a figureData governance is the set of rules and roles for a company's data: who owns each dataset, who can see it, what each number means and how long it's kept.
- Data lake, has a figureA data lake is a place where a company stores its data as files, as it arrives and untransformed, and decides later how to read it.
- Data modeling, has a figureData modeling is deciding how data is organized into tables: what each one holds, what one row means and how they connect, so questions about it get the right answer.
- Data science, has a figureData science is using data, statistics and code to answer questions and make predictions, like which customers are about to cancel or how much a store will sell next month.
- Data warehouse, has a figureA data warehouse is a database built for analysis: it brings data from many systems into one place, keeps its history and organizes it for queries.
- Database, has a figureA database stores data in connected tables, with rules for each column, so many people and programs can read and change it without stepping on each other.
- dbt (data build tool), has a figuredbt (data build tool) turns SQL SELECT queries into tables or views in your data warehouse, runs them in order and tests the results: dbt run builds the models, dbt test checks them.
- Edge caseAn input at the edge of what the logic expects, like a code that isn't in the lookup table, to see that it fails the way it should instead of returning a plausible wrong value.
- ETL (extract, transform, load), has a figureETL (extract, transform, load) copies data out of the systems where it's created, cleans and reshapes it, and loads it into a data warehouse so it can be analyzed together.
- Excel macros, has a figureAn Excel macro is a series of steps saved as code (VBA) that you can repeat with one click, like formatting, sorting and totaling a report.
- Grain (of a table)What one row of a table stands for: a person, a transaction, a planet, a person per language. Counting rows counts what you want only if you know the grain.
- INDEX MATCH, has a figure, says how to do itINDEX MATCH is two Excel functions used together to look up a value: MATCH finds where your value is, and INDEX brings back the data from another column at that spot.
- KPI (key performance indicator), has a figure, says how to do itA KPI is a number a team tracks to know whether it's doing well at what matters most, like monthly sales or orders delivered on time. It only works if it's well defined, so two people who calculate it get the same answer.
- Outlier, has a figureAn outlier is a value that sits far from the rest, like a $9,800 order among $300 ones. It can be a data-entry error or a real, important case: before you drop it, find out which.
- Pivot table, has a figureA pivot table summarizes a long table in seconds: it groups by the columns you pick and sums, counts or averages another, with no formulas.
- Power Query, has a figurePower Query is the tool in Excel and Power BI for bringing in data and cleaning it with steps that get saved: dropping columns, changing types, filtering rows. When the new file arrives, you click Refresh and every step runs again on its own.
- ReconciliationComparing a total you computed with the same total from an independent source, such as a bank statement or a register report, and explaining any difference.
- Semantic layer, has a figureA semantic layer defines each business metric, what it counts and how to calculate it, in one place that dashboards, reports and AI agents all read from.
- Slowly changing dimension (SCD), has a figureA slowly changing dimension (SCD) is a dimension table whose values change now and then. Its type says how the change is kept: overwrite the value (type 1) or keep every version (type 2).
- Spot checkTaking a few rows spread across the data (from the start, the middle and the end) and comparing them by hand with what you expect.
- SQL, has a figureSQL is the language for asking a database for data: which columns, from which table, which rows and how to summarize them. You get a table back.
- SQL join, has a figureA SQL join combines rows from two tables where a column matches, like orders with their customers. The type of join decides what happens to rows without a match.
- Star schema, has a figureA star schema arranges data for analysis: a fact table of events, like sales, in the center, joined by keys to dimension tables (customer, product, date) that describe them.
- SUMIF, has a figure, says how to do itSUMIF is an Excel and Google Sheets function that adds up only the cells that meet one condition, like one store's sales: =SUMIF(range, criteria, sum_range).
- Survivorship bias, has a figureSurvivorship bias is drawing conclusions only from what made it through a filter, like the businesses still open, without counting what fell away.
- VLOOKUP, has a figure, says how to do itVLOOKUP looks for a value in a table's first column and returns the data from another column in that row: =VLOOKUP(E2,A2:C9,3,FALSE).
- XLOOKUP, has a figure, says how to do itXLOOKUP looks for a value in one column and returns the data from the same row in another column, to the right or the left: =XLOOKUP(E2,B2:B9,C2:C9).
Percentages and averages26 words
- A/B testing, has a figureA/B testing is comparing two versions of something, like a web page or an email, by showing each one to a random half of people and measuring which does better.
- CAGR (compound annual growth rate), has a figure, says how to do itCAGR (compound annual growth rate) is the steady yearly rate that would take a value from its start to its end over a number of years: (end ÷ start)^(1 ÷ years) − 1.
- Chi-square test, has a figureA chi-square test checks whether two categories are related, like plan type and churn, by comparing the counts you got with the counts you'd expect if they weren't.
- Confidence interval, has a figureA confidence interval is the range where the whole population's number probably falls, worked out from a sample: 52% ± 5 runs from 47% to 57%.
- Cross-multiplication (rule of three), has a figure, says how to do itCross-multiplication (the rule of three) finds a missing number when two amounts change in proportion. If 3 lb cost $87, 5 lb cost 87 × 5 ÷ 3 = $145.
- Frequency table, has a figureA frequency table counts how often each value shows up (absolute frequency). Relative frequency is that count over the total: 5 of 10 = 50%.
- Interquartile range (IQR), has a figure, says how to do itThe interquartile range (IQR) is how wide the middle half of your data is: the third quartile minus the first. It tells you how spread out the typical values are.
- Mean (average), has a figure, says how to do itTo find the mean (the average), add up all the values and divide by how many there are: 15, 7, 8, 13 and 7 add up to 50, and 50 ÷ 5 = 10.
- Median, has a figure, says how to do itThe median is the middle value when you sort the data from lowest to highest: half sit below it and half above. In Excel you get it with =MEDIAN(B2:B11).
- Mode (statistics), has a figure, says how to do itThe mode is the value that shows up most; the mean is the average, and the median is the middle value. In 3, 5, 5, 8, 9: mean 6, median 5, mode 5.
- Moving average, has a figureA moving average is the average of the last few days (or months), recalculated at every point, to smooth a series and show its trend.
- Normal distribution (bell curve), has a figure, says how to do itA normal distribution is the bell shape a lot of data takes: most values near the average, fewer and fewer out to the sides, evenly on both.
- P-value, has a figure, says how to do itA p-value is the probability of seeing a result at least as extreme as yours if there were truly no effect. The smaller it is, the more surprising your result would be without one.
- Percent change, has a figure, says how to do itPercent change says how much a number moved relative to where it started: (new − old) ÷ old × 100. If sales go from $80K to $100K, they're up 25%.
- Percentage, has a figure, says how to do itA percentage says how many parts out of every 100 something is. You get it by dividing the part by the total and multiplying by 100: 45 out of 180 customers is 45 ÷ 180 × 100 = 25%.
- Percentage of a percentage, has a figure, says how to do itTo find a percentage of a percentage, multiply the two as decimals: 25% of 60% is 0.25 × 0.60 = 0.15, which is 15% of the total.
- Percentile (and quartiles), has a figure, says how to do itA percentile is the value a given share of the data falls at or below: 90% falls at or below the 90th. Quartiles are the 25th, 50th and 75th.
- Population and sample, has a figureThe population is the whole group you want to know about; the sample is the part of it you actually measure to estimate the rest.
- Probability, has a figure, says how to do itProbability measures how likely something is, from 0 (never) to 1 (certain). Favorable outcomes ÷ possible outcomes: rolling a 6 on a die is 1/6.
- Sampling, has a figureSampling is how you pick your sample. The four types: simple random, systematic (one in every so many), stratified (random within each group) and cluster.
- Simpson's paradox, has a figureSimpson's paradox is when a trend that shows up in every group reverses once the groups are combined, because each side has a different mix of groups.
- Standard deviation, has a figure, says how to do itStandard deviation measures how far data lands from the average: with an average of 30 and an SD of 2, most values fall from 28 to 32. In Excel: =STDEV.S.
- t-test, has a figureA t-test checks whether the difference between two averages, like the average basket before and after a change, is likely real or could be chance.
- Variance, has a figure, says how to do itVariance measures how spread out data is: it averages the squared distances from the mean. Its square root is the standard deviation.
- Weighted average, has a figure, says how to do itA weighted average gives each value a weight for how much it counts, instead of treating them all the same: multiply each value by its weight, add them up and divide by the sum of the weights.
- Z-score, has a figure, says how to do itA z-score says how many standard deviations a value sits from the mean: z = (value − mean) ÷ standard deviation. With mean 50 and SD 10, a 60 has z = 1.
Python7 words
- Google ColabGoogle Colab is a free Google service that runs Jupyter notebooks in your browser, with nothing to install. pandas and Altair come already installed.
- Jupyter notebookA Jupyter notebook is a file (ending in .ipynb) that mixes code, its results and notes in cells. You run one cell at a time and see its output right below it.
- PackageA package is a bundle of code someone else wrote that you install to give Python new tools. pandas and Altair are packages.
- pandaspandas is a Python package for working with tables of data, called DataFrames: loading them, filtering rows, picking columns and summarizing them.
- Tuple, says how to do itA tuple is a sequence of values in parentheses, like (16, 5). It works like a Python list that stays the same after it's made.
- Unit testGiving a piece of logic (a formula, a function) an input whose correct output you already know, and checking that it returns that.
- Virtual environmentA virtual environment is a private set of Python packages for one project. What you install there stays there, so each project keeps the versions it needs.
Your computer12 words
- Algorithm, has a figureAn algorithm is a series of precise steps that, followed in order, solve a problem, like a recipe, an Excel formula or a search.
- API, has a figureAn API (application programming interface) is a set of rules that lets one program ask another for data or actions, like your script asking a store for its orders.
- Command PaletteThe Command Palette is VS Code's search box for every command it has. Open it with Ctrl+Shift+P (Cmd+Shift+P on a Mac), type a few words and press Enter.
- CSV, says how to do itA CSV (comma-separated values) is a plain-text file that holds a table: one row per line, with commas between the columns.
- ExtensionAn extension is an add-on you install in an app like VS Code to give it new abilities, such as running Python or opening notebooks.
- JSON, says how to do itJSON is a text format for structured data: labels and values inside curly braces, lists inside square brackets. Programs use it to pass data to each other.
- PathA path is a file's address on your computer: the folders you go through to reach it, ending in the file's name, like projects/python-for-altair/life_expectancy.csv.
- Python, has a figurePython is a free programming language with short, readable code. People use it to clean and analyze data, make charts, automate tasks and build AI agents.
- Script, has a figureA script is a short program, saved in a text file, that does one task from start to finish every time you run it, the same way each time. It doesn't think or guess: it reads what you give it, does its steps and hands back a result or an error.
- TerminalA terminal is a window where you type commands for your computer instead of clicking. You type a line, press Enter, and the computer answers in text.
- VS CodeVS Code (Visual Studio Code) is a free code editor from Microsoft. With its Python and Jupyter extensions, it runs notebooks right inside the editor.
- Working directoryThe working directory is the folder a program treats as "here". A file name with no folders in front of it, like life_expectancy.csv, is looked for in that folder.
See these words at work
Course pages mark terms like these where they come up, and your Maestro explains them in the sentence you are reading.
Free courseClaude Code for Beginners: Your First Session with DataClaude Code handles the basic analysis well. You know what to ask it, how to check what it did, and which number you put your name on.Start reading