Career roadmap

How to become a data engineer in 2026.

A practical, no-fluff guide. No "buy my $2k course". Just the skills that actually get hired, the order to learn them in, and the projects that prove you can do the job.
Start the roadmap See all 14 courses

Step 1 — Learn the language: SQL

Every interview, every code review, every on-call incident touches SQL. Don't skim it. Get to the point where window functions, CTEs and joins on three tables are reflex, not effort.

If you can answer "what's the 7-day rolling average per user, ignoring days with zero events?" in a clean SQL query under 5 minutes, you're past step 1.

Step 2 — Pick up Python (the data way)

You don't need to write Django apps. You need to: read JSON from APIs, parse files, push data into a warehouse, write a function someone else can read. Pandas/Polars for transforms, requests/httpx for HTTP, pydantic for schemas. That's the kit.

Step 3 — Containers and infrastructure

Docker is the dividing line between "tutorial dev" and "real engineer". Learn to write a Dockerfile, run docker-compose, debug a container that won't start. Then add Terraform — every modern team provisions cloud with code, not clicks.

Step 4 — A warehouse + a transformation layer

Pick one warehouse (Snowflake, BigQuery, or Redshift) and learn it deeply. Then learn dbt on top — sources, staging, marts, tests, snapshots. This is the layer most companies will hire you to own first.

Step 5 — Orchestration

Pipelines that run once are toys. Pipelines that run every hour with retries, SLAs, alerting and lineage are the job. Learn Airflow (still the most-asked) or Dagster (where the modern stack is going).

Step 6 — Distributed processing & streaming

Once batch is comfortable, scale up. Spark for big batch jobs, Iceberg for the table format, Kafka for streams. You don't need them on day one — but every senior data engineer knows them.

Step 7 — Build one real project, end-to-end

This is what unlocks interviews. Build something like:

  • Ingest a public API (weather, stocks, GitHub events) into S3 or a warehouse.
  • Transform with dbt into clean marts.
  • Schedule with Airflow / Dagster, with tests and alerts.
  • Expose one dashboard or one API on top.
  • Containerise it and put the whole thing in a public GitHub repo.

One repo like that gets more interviews than three certificates.

How DataForge fits in

DataForge walks you through steps 1–6 in 14 gamified courses with daily 5-minute exercises. You still build the project (step 7) yourself — but you'll have all the muscle memory and reference patterns to do it without getting stuck for a week on a missing import.

FAQ

How long does it take to become a data engineer?
From scratch, with 30 minutes of practice a day, most learners reach an entry-level data engineer level in 4–6 months and a mid-level (independent on a real platform) level in 12–18 months.
Can I become a data engineer without a degree?
Yes. The job market in data engineering is one of the most skill-driven in tech. A clean GitHub with one real project (ingestion + warehouse + dbt + dashboard) outweighs most degrees.
What programming language should a data engineer know?
SQL is non-negotiable. Python is the default for ETL, orchestration and data tooling. Scala or Java come later if you go deep into Spark or Flink. Bash and YAML are everyday.
Do data engineers need to know machine learning?
No, but understanding what ML pipelines need from you (feature stores, training data, freshness, lineage) makes you instantly more useful. Don't start with ML — finish the data engineering fundamentals first.
Data engineer vs analytics engineer vs ML engineer — which one?
Analytics engineers focus on dbt and the warehouse. ML engineers focus on training and serving models. Data engineers own the platform that feeds both. If you like building systems, pick data engineering.

Ready to start?

7 days free. Then less than a coffee per month.

Start the roadmap