Guide - Data

The modern data stack, explained through listings

What each layer of the data stack does, and how often remote data listings ask for it.

Grounded in 4,022 remote positionsData captured October 2026

"Modern data stack" describes a set of cloud tools that together move raw data into a warehouse, transform it into models and serve it to analysts and applications. Data engineering listings in the sample name its components consistently, which makes the stack unusually learnable.

The layers

  • Languages: SQL (98 of 275 data engineering positions) and Python (106) are the base.
  • Storage and compute: a cloud warehouse or lakehouse such as Snowflake (60), Databricks (48) or BigQuery (34).
  • Transformation: dbt (48) turns raw tables into tested, documented models using SQL.
  • Orchestration: Airflow (40) or Dagster schedules and monitors the pipeline.
  • Processing at scale: Apache Spark and PySpark for large or complex transformations.
  • Serving: BI tools such as Power BI, Tableau and Looker.

ETL or ELT?

Classic ETL transforms data before loading it; ELT loads raw data first and transforms it inside the warehouse, which is how dbt works. Listings name ETL far more often (51 positions against 11 for ELT), but often use it loosely for any pipeline work. See ETL vs ELT.

How specific the stack is

These tools rarely appear outside data roles. 79% of Snowflake demand and 87% of dbt demand falls in data engineering. That makes them strong differentiators inside the family and weak signals outside it.

The data engineering learning path orders these skills into a sequence, and the learning priority includes a portfolio project that exercises the whole stack.