Skip to content

Data team roles#

Optional

The jobs people do on a data platform, such as data engineer, analytics engineer, analyst, data scientist, ML engineer, platform engineer and data product owner, what each is responsible for, and where they overlap.

Before reading

Data products and data mesh

Why it matters#

Job titles in data vary between companies, and the same title can mean very different work. When responsibilities are unclear, work falls between roles: nobody tests the gold tables, nobody owns the pipeline a data scientist wrote, or two people build the same metric twice. Knowing the usual split of work helps teams decide who does what, and helps readers of this wiki see which pages matter most for their own role.

How it works#

Roles follow the flow of data through the platform, roughly from left to right.

The platform engineer builds and runs the shared platform: storage, catalogs, query engines, orchestration, access control, CI and monitoring. Their customers are the other data roles. In a data mesh, this is the self-serve platform team.

The data engineer brings data in and makes it reliable: ingestion from databases, APIs and streams, change data capture, raw and clean layers, deduplication, schema changes and performance. They work mostly in the lower layers and care about completeness, correctness and cost.

The analytics engineer turns clean data into models people can use: dimensional models, gold tables, metric definitions in the semantic layer, tests and documentation, usually in dbt and SQL. The role appeared with tools like dbt, which let people with SQL skills build tested, version-controlled transformations.

The data analyst answers business questions with data: reports, dashboards, ad hoc analysis. They are the main consumers of gold tables and the semantic layer, and often the first to notice when numbers look wrong.

The data scientist investigates questions that need statistics or models: experiments, forecasts, segmentation, causal analysis. They work in notebooks and Python, reading clean and gold data, and hand over results that become metrics or models.

The ML engineer turns models into production systems: training pipelines, feature stores, registries, serving and monitoring. Their focus is reliability and scale of models, where the data scientist's is their quality.

The data product owner is accountable for a data product as a product: what consumers need, what it promises, and what changes next. This often overlaps with the data owner role in governance.

flowchart LR
  SRC[Sources] --> DE[Data engineer:<br/>raw and clean layers]
  DE --> AE[Analytics engineer:<br/>gold models, metrics]
  AE --> DA[Analyst:<br/>dashboards, analysis]
  AE --> DS[Data scientist:<br/>experiments, models]
  DS --> MLE[ML engineer:<br/>models in production]
  PE[Platform engineer:<br/>shared platform] -.-> DE
  PE -.-> AE
  PE -.-> MLE

In practice, roles overlap and combine. A small team may have two people who do everything; a large company may split each role further. What matters is that each layer and each published asset has someone responsible for it, and that hand-overs, such as a data scientist's prototype becoming a production pipeline, are explicit.

These are job roles. Governance roles, such as data owner and steward, describe decision rights and are covered in Governance roles; one person often holds both kinds.

Key terms#

  • Platform engineer Builds and operates the shared data platform.
  • Data engineer Builds ingestion and the reliable lower layers.
  • Analytics engineer Builds tested models and metrics for consumers.
  • Data analyst Answers business questions with reports and analysis.
  • Data scientist Uses statistics and models to investigate and predict.
  • ML engineer Puts models into production and keeps them running.
  • Data product owner Accountable for a data product's value to its consumers.

Example#

Who builds what for the shop's daily sales report:

Work Role
Kafka, catalog, Trino cluster, dbt CI Platform engineer
CDC from the orders database into raw_orders, clean_orders merge Data engineer
fct_daily_sales, the revenue metric, tests and contract Analytics engineer
The finance dashboard and weekly commentary Data analyst
The demand forecast built on feat_product_daily Data scientist, then ML engineer
Deciding what fct_daily_sales promises to finance Data product owner (head of sales operations)

Pitfalls#

  • Titles without responsibilities. Agree who owns each layer, not only what people are called.
  • Data scientists running production pipelines alone. Prototypes need an explicit hand-over to engineering.
  • No analytics engineering. Analysts build their own models in BI tools, and definitions diverge.
  • Platform teams building for themselves. The platform's users are the other roles; ask what they need.

Learn more#

Governance roles Medallion layers: bronze, silver and gold dbt The ML lifecycle and MLOps