AI and data roles are the jobs that show up in job ads under English titles: Data Analyst, Data Scientist, Data Engineer, ML Engineer, AI Engineer and others. Each one covers a part of the chain: gathering the data, analyzing it, building models, running them in production or building products with AI.
A question from the app. Pick an answer.
You need a chat over the manuals, using an existing model. Which role usually does it?
Builds products on models that already exist: connects them, feeds them the documents and checks the answers. It overlaps with ML Engineer, but this is its core.
Builds the pipelines that fetch, clean and store the data. Connecting a model and checking its answers is another job.
Answers business questions with data: dashboards, reports, queries. Building a chat on a model is another job.
Answers business questions with data: why sales dropped in March, which product sells best in each city. Queries the data (often with SQL), compares periods and tells the story with charts and dashboards anyone can follow. Usually works with SQL, spreadsheets like Excel or Google Sheets and dashboard tools like Power BI, Tableau or Looker.
Often confused with Data Scientist. The usual difference: the analyst explains what happened; the Data Scientist builds models to predict what will happen.
Looks for patterns in the data and tries out machine learning models: one that forecasts sales, another that spots which customers are about to leave. Runs experiments, measures how well each model does and explains the results. Usually works with Python, pandas, scikit-learn, notebooks like Jupyter and a good deal of statistics.
Confused with Data Analyst on one side and ML Engineer on the other: the Data Scientist usually gets as far as a model that works in tests; running it for thousands of users is more engineering work.
Builds the pipelines: processes that pull data from several systems, clean it and store it in one place, on their own and on schedule (every night, say). Without that work, the other roles use scattered or stale data. Usually works with SQL, Python, orchestrators like Airflow, tools like dbt or Spark and cloud data warehouses like BigQuery or Snowflake.
Confused with Data Scientist because both touch the same data. The tell: the Data Engineer moves and organizes data, and does not need a model to do it.
Trains, adapts and ships models, in-house or pretrained: making sure the model that predicts delays answers fast, handles the traffic and can be updated. Writes more production code than a Data Scientist. Usually works with Python, PyTorch, models and libraries from Hugging Face, Docker and cloud services.
Confused with AI Engineer, and the tasks overlap. The tell: the ML Engineer looks after the model (training, adapting, running it); the AI Engineer, after the product built with AI.
MLOps is more a practice than a profession: getting a model into production in an orderly way and keeping it healthy. It covers versioning data and models, automating training and releases, and watching whether accuracy drops when the data changes (another season, other customers) so it can be retrained in time. Example tools: MLflow, Kubeflow, Docker and monitoring systems.
In job ads it shows up as a role (MLOps Engineer) or as a task inside ML Engineer. Confused with DevOps: it is the same idea of automating and monitoring, applied to models that can get worse without the code changing.
Builds products on models that already exist: a chat that answers from the company manuals, an assistant that sorts emails, an agent that fills in forms. Connects the model through an API, gives it the right documents (with RAG, for example) and checks the answers with evals. Usually works with model APIs like Claude, GPT or Gemini, embeddings, vector databases and backend code.
Confused with ML Engineer (see above) and with AI-Native Engineer: the AI Engineer builds products with AI; AI-Native Engineer has no uniform definition (see below).
It has no uniform definition: it is a job ad label, and each company uses it its own way. Some ads focus on coding with AI every day, with assistants and coding agents like Claude Code, Codex or Cursor, and reviewing what they do; others, on building AI products and agents.
Confused with AI Engineer and with "someone who trains large models from scratch", which is the job of a few research teams. Since it is not a fixed role, read the tasks in the ad.
Titles split tasks; people often do not. At a small company, whoever gathers the data also analyzes it or builds the model, and one job can cover the whole chain. At a big one, roles tend to be more split.
So look at the tasks, not the title. "Build dashboards for management" is analyst work even if the ad says Data Scientist. "Train and deploy models" is ML Engineer work even if it says AI Engineer. If the tasks are unclear, ask what a typical week would look like.