Data Engineer
Data engineers build the pipelines and storage that everyone else's dashboards, models, and metrics depend on. This page covers what the work actually is, what it pays, the specific tools that get you hired, and why most of the job is reliability and cost rather than cleverness.
Typical Pay (US)*
$130kmedian** AI-estimated from general U.S. labor-market patterns — not measured data from the U.S. Bureau of Labor Statistics or any official source. Real pay varies widely by location, employer, experience, and timing.
Outlook
This is one of the safer bets in data: Mochivia's estimate for the nearest O*NET occupation is about 6.5% annual growth with an estimated automation-exposure score of 0.25, and warehouse modernization plus the data hunger of AI systems keeps demand rising. AI genuinely eats parts of the job — boilerplate connector code, transformation scaffolding, DAG skeletons, and the first draft of a schema migration are all faster to prompt than to type. What it does not eat is the decision-heavy core: choosing the storage layout that keeps a query cheap, designing schema-evolution contracts across teams, diagnosing why a distributed job is slow, and owning the pager when data is late. The role shifts from writing pipelines toward designing and operating them, which raises the bar on systems knowledge and lowers the value of tool trivia.
What does a Data Engineer do?
A data engineer makes data arrive — on time, complete, and in a shape someone can query. A payments team emits events to Kafka; you write the ingestion job that lands them in S3, the Spark or dbt-adjacent transform that partitions them by day, the Airflow or Dagster DAG that runs it at 03:15, the freshness check that pages someone when the 03:15 run produces zero rows, and the backfill script for the week the upstream schema changed silently. Then you get a Snowflake bill and spend two days figuring out which query is costing $4,000 a month. That is the actual texture of the job: plumbing, contracts, monitoring, and money.
The distinction from the neighboring roles is sharp once you see it. A data engineer owns everything up to and including the raw and lightly-cleaned tables in the warehouse — ingestion, orchestration, storage layout, streaming, infrastructure. An analytics engineer picks up from there and builds the modeled, tested, business-meaningful tables in dbt. A data analyst queries those models to answer questions. A data scientist does inference and modeling on top. An MLOps engineer does the same reliability work for model training and serving instead of tables. If someone asks "why is the dashboard empty," the answer usually lands on the data engineer.
O*NET 28.3 maps this most closely to occupation 15-1243, Database Architects, which lists "Big Data Engineer" and "Cloud Architect" among its aliases and reports a median around $135,980 with a band from roughly $103,000 to $175,000 across about 62,000 employed workers. It also scores unusually well on the two numbers that matter for a career bet: about 6.5% estimated annual growth and a low automation-exposure score of 0.25, placing it in the 96th percentile for momentum. The federal rationale is blunt — warehouse-modernization demand drives the growth, and the judgment-heavy design work is among the least automatable in the data stack.
The role suits people who get satisfaction from systems that stay up. The best data engineers are, temperamentally, backend engineers who happen to care about data: they think about idempotency, schema evolution, retries, and cost per terabyte scanned. It does not suit people who came to data for the analysis — you will rarely be the one interpreting the numbers, and the interesting business question usually belongs to someone downstream. The clever distributed-systems trick is maybe 10% of the work. The other 90% is making sure last night's run finished and did not cost a fortune.
A day in the life
- Start with the pipeline alert channel: three DAGs failed overnight, one on an upstream schema change and two on a transient S3 timeout
- Write a backfill for six days of Kafka events that were dropped when a consumer group rebalanced badly
- Review a colleague's pull request adding a Snowflake table, and push back on the partition key because it will full-scan every query
- Track down which dbt-adjacent transform quadrupled last month's compute bill, then rewrite it to prune partitions
- Sit with an analytics engineer to agree on a data contract so a source-column rename stops breaking downstream models silently
- Add freshness and row-count tests to a pipeline nobody has monitored since the person who wrote it left
- Spend an hour on the migration plan moving batch ingestion for one domain from cron scripts to Dagster
How to become a Data Engineer
- 1
Learn Python and SQL to production depth
~4 monthsNot scripting depth. You need Python you would put on call — typed, tested, packaged — and SQL where you can read a query plan and explain why it is slow.
- 2
Learn one cloud warehouse properly
~2 monthsPick Snowflake, BigQuery, or Redshift and learn its storage model, partitioning, clustering, and pricing mechanics. Cost tuning is a hiring signal that separates candidates fast.
- 3
Build a real orchestrated pipeline end to end
~3 monthsIngest a live public API on a schedule with Airflow or Dagster, land it in a warehouse, transform it, add data-quality tests, and make failures visible. Ship it and keep it running.
- 4
Add distributed processing and streaming
~3 monthsLearn Spark for batch at scale and Kafka for event streams — enough to reason about partitions, shuffles, consumer groups, and exactly-once semantics rather than just running examples.
- 5
Pick up the infrastructure adjacent skills
~2 monthsDocker, Terraform, CI pipelines, and basic Kubernetes. Data platform work sits on infrastructure, and the engineers who can provision their own resources move much faster.
- 6
Enter through an adjacent seat if you are not hired directly
~6 monthsBackend engineering, analytics engineering, or a data-analyst role at a company with a small platform team all convert well. No degree is formally required; a running pipeline in production is worth more than a certificate.
Skills that matter
Learn the actual skills
Mochivia's structured roadmap walks you from fundamentals to job-ready — 15 minutes a day.
See the Roadmap