MLOps & Data Engineering

Module 17 of 18

Module 17: Glossary of Data & MLOps Terms

4 min read764 words
What you'll learn
Recognize the key terms used across this trackExplain each in simple languageUse this page as a quick reference

"One page, the whole vocabulary. Once these feel familiar, data engineering job posts and docs read like plain English."

Level: Reference · Time: ~browse · Prerequisites: none

Learning Objectives

By the end of this module, you will be able to:

  • Recognize the key terms used across this track
  • Explain each in simple language
  • Use this page as a quick reference

Use this the way you'd use a phrasebook in a new country: skim it once now so the words feel familiar, then come back whenever a term shows up in a job post, a colleague's Slack message, or a tool's documentation. You don't need to memorize it — you need to recognize the terms and know roughly where each one sits in the bigger picture. The grouping below mirrors the order of the track, from raw data basics up through MLOps.

A quick note on how these terms relate: the vocabulary is layered, not flat. Data-basics words (schema, batch, streaming) describe the raw material; storage words (warehouse, lake, index) describe where it rests; processing words (ETL, Spark, Kafka) describe how it moves; orchestration and ops words (DAG, Docker, cloud) describe how it's coordinated and run; and MLOps words (deployment, drift, registry) describe operating models on top of all that. If a term ever feels disconnected, ask which of these layers it belongs to — that placement usually makes its meaning click.

Data Basics

TermPlain meaning
Structured dataRows and columns (tables)
Semi-structured dataFlexible formats like JSON or XML with some shape
Unstructured dataImages, text, audio — no fixed shape
SchemaThe defined structure/types of data
Schema driftWhen data's shape changes unexpectedly and breaks pipelines
BatchProcessing data in scheduled chunks
StreamingProcessing each event as it arrives
PipelineAutomated flow that moves/transforms data

Storage & Databases

TermPlain meaning
Relational databaseData in linked tables, queried with SQL
SQLThe language for querying databases
NoSQLFlexible non-tabular databases
JoinCombining rows from tables on a shared key
IndexStructure that speeds up queries
Data warehouseSystem built for large-scale analytics (OLAP)
OLTP vs OLAPRunning the app vs analyzing the data
Data lakeCheap storage for raw data of any format
LakehouseLake + warehouse features combined
Data swampA disorganized, untrustworthy lake
Columnar storageStoring by column for fast analytics
Object storageCheap, vast cloud file storage

Processing & Movement

TermPlain meaning
ETL / ELTExtract-Transform-Load / Load-then-Transform
IdempotentSafe to run repeatedly, same result
Incremental loadProcessing only new data, not everything
SparkEngine for distributed big-data processing
Distributed processingSplitting work across many machines
Lazy evaluationBuilding a plan, running only on an action
KafkaDurable backbone for streaming events
Producer / topic / consumerSends / stream of / reads events

Orchestration & Ops

TermPlain meaning
OrchestrationCoordinating tasks in order, on schedule
DAGDirected Acyclic Graph of tasks (no loops)
BackfillRe-running a workflow over past dates to fix history
AirflowPopular Python orchestration tool
ContainerCode packaged with its whole environment
Image / containerBlueprint / running instance
DockerThe common container tool
KubernetesOrchestrates many containers at scale
Cloud (AWS/Azure/GCP)Rented, on-demand infrastructure
ElasticityScaling resources up/down with demand

MLOps

TermPlain meaning
MLOpsDeploying and maintaining models in production
Deployment / servingMaking a model usable via batch or API
Training/serving skewData prepped differently in training vs serving
CI/CD/CTAutomated test / deploy / continuous training
Experiment trackingRecording each training run's data & results
Model registryVersioned catalog of models (staging/production)
DriftData/accuracy diverging from training over time
Data governanceRules for ownership, access, privacy, retention
LineageWhere data came from and how it changed
ObservabilityBeing able to see a system's health from its signals

Try this: Pick any five terms above that felt fuzzy the first time you met them. Cover the right-hand column and explain each in your own words. The ones you can't explain plainly are exactly the concepts worth revisiting in their module — recognition is good, but being able to teach a term back is how you know it's yours.

Key Takeaway: These terms span the whole track — data basics, storage and databases, processing and movement, orchestration and operations, and MLOps. Bookmark this page as your quick reference while you build pipelines and ship models.

Further Learning

Part of "MLOps & Data Engineering." Original content for this learning platform.