Data and AI vocabulary without the jargon tax. Everything lives on this one page — jump to a topic, or scroll the full glossary.
Data Governance ETL / ELT Clean rooms MDM KPIs Data literacy · AI LLM Hallucination Prompt RAG Embeddings Vector DBs Agents Eval
Data
Data Governance
The strategy, policies, standards, and practices for managing data as a valued asset across its lifecycle. Strong governance improves data quality, creates clear ownership, and supports compliance — so people can trust the numbers they use to decide.
ETL vs ELT
Two patterns for moving data between systems. ETL (extract, transform, load) cleans and reshapes data before it lands in the warehouse. ELT loads first and transforms in the warehouse, which often fits modern cloud platforms better.
Data Clean Rooms
Shared analysis environments designed so organizations can collaborate on data without freely exchanging raw personal information. Useful when privacy rules matter but two parties still need joint insight.
Master Data Management (MDM)
Creating and maintaining a single, trusted view of core business entities — customers, products, suppliers — across fragmented systems.
KPIs (Key Performance Indicators)
Metrics that track progress toward goals that matter. Good KPIs are few, owned, and tied to decisions. Bad ones are vanity counts that look busy but never change what the team does next week.
Data Literacy
The ability to read, work with, analyze, and argue from data — a skill for whole teams, not only analysts.
AI
LLM (Large Language Model)
AI systems trained on large amounts of text that can generate and understand language — the class of tool behind products like ChatGPT. Useful for drafting, summarizing, exploring ideas, and coding help. Still need human review for accuracy, privacy, and bias.
Hallucination
When a model produces a fluent, confident answer that is wrong or made up — fake citations, invented metrics, SQL that can’t run. Validate outputs against source data and schema before anything goes to a stakeholder.
Prompt (and context window)
The instructions and information you give a model. Clear prompts state the goal, format, constraints, and what “good” looks like. The context window is how much text the model can consider at once.
RAG (Retrieval-Augmented Generation)
A pattern where the system first finds relevant snippets from your docs or data, then asks the model to answer using that material. Goal: fewer pure guesses, more answers grounded in sources you control.
Embeddings
Numeric representations of text (or other data) that capture meaning so “similar” items sit near each other in math space. Used for semantic search, recommendations, and many RAG pipelines.
Vector Databases
Databases optimized to store embeddings and search by similarity — “find things like this,” not only exact keywords. Common building blocks for semantic search and AI apps.
Agents (AI)
Setups where a model can plan steps and call tools (search, code, databases) rather than only reply once in chat. In practice: narrow tools, clear permissions, logging, and a human in the loop for anything that changes production data.
Eval (Evaluation)
Checking whether AI outputs are good enough for a real task — accuracy, format, safety, cost. If you can’t define “good,” you can’t improve the system; shipping without evals is hoping in production.