Interactive learning journey · 10 modules

One system.
Many moving parts.

Data engineering moves data from operational reality to trusted decisions: securely, reliably, and at a sustainable cost. Tools change; the responsibilities endure.

Level Foundation → IntermediateModules 10 deep-dive topicsFormat Read · interact · reflectAuthor Tarek Atwan @ ML_LAB

Your learning contract

By the end, you can explain the system, not just name tools.

Prerequisites: basic SQL, familiarity with files and databases, and curiosity about how data reaches a dashboard.

  • Trace a record from source to consumption.
  • Compare batch, CDC, and streaming patterns.
  • Separate storage, file, and table formats.
  • Explain Bronze, Silver, and Gold by purpose.
  • Design for reruns, late data, and failures.
  • Choose an architecture from requirements and trade-offs.
  • Compare Inmon, Kimball, and Data Vault modeling.
  • Explain Lambda vs. Kappa streaming architectures.

The constant pattern

Tools change.
The data journey does not.

Switch platforms to see the same responsibilities expressed through different services. Select a layer to zoom from the architecture down to engineering decisions.

Active viewConceptualSame responsibilities · different implementation

04 · Trust

Silver · Conformed

Make data consistent, valid, and reusable across domains. Silver is where engineering quality becomes organizational trust.

  • Deduplicate, validate, and quarantine
  • Standardize dates, currencies, units, and codes
  • Resolve identities and conform shared entities
Architect's question: Is this data trustworthy enough to reuse without relearning every source?
Across every layer:SecurityGovernanceQualityLineageObservabilityFinOps

Learning companion · 10 modules

Architecture is the map.
Practice makes the engineer.

Each module is a self-contained deep dive. Work through them in order for the best learning progression, or jump to any topic you need to review.

0 of 10 modules complete
Module 01

Foundations & Requirements

Start with the decision, not the tool
  • The data engineer's role & responsibilities
  • OLTP vs. OLAP systems
  • SQL and data profiling fundamentals
  • Latency, volume, SLAs & constraints
  • Understanding data consumers
Module 02

Sources & Ingestion

Extract without harming sources
  • Batch, incremental & full loads
  • Change Data Capture (CDC)
  • Event-driven ingestion
  • Schema drift & data contracts
  • Tools: Airbyte, Fivetran, dlt
Module 03

Storage & Formats

Where bytes live and how engines read them
  • Row-based vs. columnar storage
  • Parquet, JSON, Avro, ORC
  • Data lake, lakehouse & warehouse
  • Delta Lake, Iceberg, Hudi
  • Partitioning & the small files problem
Module 04

Transform & Quality

Raw data rarely arrives ready to trust
  • Bronze → Silver → Gold progression
  • Profile, deduplicate, validate, standardize
  • Data quality & testing frameworks
  • Libraries: Pandas, Polars, DuckDB, PySpark
  • Tools: dbt, Great Expectations, Soda
Module 05

Data Modeling

Clean data is not yet business-ready data
  • Bill Inmon: 3NF top-down approach
  • Ralph Kimball: Star schema, facts & dimensions
  • Data Vault 2.0: Hubs, links & satellites
  • ETL vs. ELT, schema on read vs. write
  • Slowly changing dimensions (SCD)
Module 06

Streaming & Real-Time

Reason about time, state, and ordering
  • Event streams & event brokers
  • Lambda vs. Kappa architecture
  • Event time vs. processing time
  • Watermarks & delivery semantics
  • Tools: Kafka, Flink, Kafka Streams
Module 07

Orchestrate & Operate

A pipeline is not finished when it works once
  • DAGs, retries & backfills
  • Idempotency & SLIs/SLOs
  • Git, CI/CD & version control
  • Docker & Infrastructure as Code
  • Tools: Airflow, Dagster, Prefect
Module 08

Govern, Secure & Observe

Governance is not a gate at the end
  • PII, masking & encryption
  • MDM & identity resolution
  • Lineage, catalog & data contracts
  • RBAC, ABAC & least privilege
  • Observability, SLIs & compliance
Module 09

Tools & Technology Landscape

Know the ecosystem, pick the right tool
  • Databricks & Snowflake deep dive
  • Apache Spark / PySpark ecosystem
  • dbt, Airflow & the modern data stack
  • Cloud platform comparison
  • When to use what: decision frameworks
Module 10

Scale, Cost & Architecture

Start with requirements, earn the complexity
  • Architecture decision framework
  • Volume, velocity, variety trade-offs
  • FinOps & total cost of ownership
  • The purpose: analytics, ML & AI
  • Data engineers and the people bridge

The modern data stack

Where each tool fits
in the pipeline.

Tools change constantly, but responsibilities don't. This map shows leading tools organized by what they do in the data journey. Highlighted blue tools represent primary industry standards covered in detail throughout the 10 modules.

Languages
SQLPythonScalaJava
Ingestion
AirbyteFivetrandltMeltanoAWS DMSAzure Data Factory
Processing
Apache Spark / PySparkPandasPolarsDuckDBAWS GlueDataproc
Transformation
dbtDataformSQLMesh
Streaming
Apache KafkaApache FlinkKafka StreamsKinesisPub/SubEvent Hubs
Orchestration
Apache AirflowDagsterPrefectKestraStep Functions
Platforms
DatabricksSnowflakeBigQueryRedshiftFabricSynapse
Table Formats
Delta LakeApache IcebergApache Hudi
Data Quality
Great ExpectationsSodaMonte CarloElementary
DevOps & IaC
Git / GitHubDockerTerraformGitHub ActionsCI/CD

Learning companion · Connected glossary

Find the term.
Reconnect it to the system.

Definitions are more useful when students can place a concept in the architecture and explain what breaks when it fails.