Data and AI roles
A reporting need, difficulty accessing data and a model to put into production do not call for the same hire. The Data and AI family brings together the roles that make this work possible and useful to teams.
Connect data with a concrete use
The Data Engineer builds the systems that collect, store and make data available. The Data Scientist uses and interprets it to answer a question or inform a decision. When models are used in production, work on them continues through deployment and monitoring their operation.
These responsibilities need to connect with the people who will use the results. Before defining a role, state the business question, identify the accessible data and clarify the decision or service expected. Agree with Engineering on responsibility for the production system and with the teams using it on how to assess its usefulness.
Understanding the different roles
Make data accessible and usable
Data engineering organises data collection, storage and availability. To distinguish your need for architecture, engineering or analytical transformation, describe the data to connect, its users and the quality problems to address. Job titles then help explore this remit in more detail.
- Data engineer
- Data architect
- Analytics engineer
Use data to inform a question
The Data Analyst produces analyses and dashboards, among other outputs, to make data useful for decision-making. The Data Scientist also develops machine learning models to address problems. The boundary is not rigid: an analyst may build models. Clarify the place of analysis, presenting findings and model development in the role.
- Data analyst
- Data scientist
Move from experimentation to monitored operation
MLOps practices connect testing, deployment, automation and model monitoring, among other activities. They help distinguish experimental work from operational responsibilities. A project may involve several specialists: specify who works on the model, its integration and its monitoring.
- ML engineer
- MLOps engineer
- LLMOps engineer
- AI engineer
Define specific AI mandates
For a need involving research, client integration or Data and AI leadership, describe the deliverables and the scope of decisions. Emerging titles need the same clarity: identify the system involved, its users and the evaluation criteria before choosing a job title.
- Chief Data / AI Officer
- Forward deployed engineer
- Prompt engineer
- AI researcher
Job profiles in this family
- Chief Data / AI OfficerLeads the company's data and AI strategy: governance, platform, use cases and scaling to production.
- Data analystA Data analyst prepares and analyses data to help business teams understand their activity and make informed decisions.
- Analytics EngineerBuilds the data transformation layer: clean, tested, documented dbt models that turn raw data into a usable asset.
- Data EngineerBuilds and operates the pipelines that collect, move and expose data at scale: ingestion, warehouse, orchestration.
- Data ArchitectDesigns the overall data architecture: models, flows, governance and platform choices, so data serves the company over time.
- Data ScientistDesigns the statistical and machine learning models that turn your data into predictions and measurable optimizations.
- ML Engineer (Machine Learning Engineer)A software engineer specialized in machine learning: designs, trains and integrates models into production systems.
- MLOps EngineerIndustrializes and operates ML models in production: pipelines, continuous deployment, monitoring and reliability.
- LLMOps EngineerOperates large language models in production: evaluation, inference cost, observability, RAG and reliability of LLM-based systems.
- AI EngineerBuilds generative-AI products and features: agents, RAG, LLM integrations — with a software engineer's rigor.
- Forward Deployed EngineerAn engineer deployed at the customer: builds on-site the solutions that turn your product into concrete value for each strategic account.
- Prompt EngineerOptimizes language-model behavior through systematic design, evaluation and iteration of prompts and system instructions.
- AI ResearcherAdvances the state of the art: new architectures, training methods, evaluation — and turns research into product advantage.
Choose according to the problem you face
Two teams use conflicting figures
If a metric changes depending on its origin, examine definitions, transformations and sources before requesting a new dashboard. Explicitly assign responsibility for resolving these discrepancies and validating the resolution with users.
Data is difficult to retrieve
If analysis depends on manual extracts or unavailable data, describe the access and flows that need to be built. This need may point towards data engineering rather than a role mainly responsible for interpretation.
A model works in a demonstration
Before arranging its real-world use, define the deployment conditions, tests and monitoring expected. For a predictive model, monitoring its performance may lead to another iteration. Assign this responsibility rather than treating the demonstration as the end of the project.
An AI feature needs to become part of a product
If the need concerns a service used by clients, clarify the interactions with existing software and the failure scenarios to handle. Involve product and technical leads in defining the role and evaluating the result.
Define the resources as well as the outcome
Check the data available
List accessible sources and known restrictions. If preparing them is part of the role, make this work visible; otherwise, identify the team responsible.
Define evaluation around use
Clarify how users will examine the results. A technical metric can help monitor a model, but the brief must also describe the service expected and the consequences of an error.
Assign responsibility for follow-up after delivery
Name those responsible for the data, the system and further changes. Plan how feedback from use will trigger a correction, further analysis or another experiment.
Frequently asked questions
How do you scope a Data assignment when the business question is still unclear?
Plan an initial scoping stage with an identified user. Ask them to state the decision they are trying to make, the information they lack and a deliverable that can test the usefulness of the work. Limit this initial scope before extending the assignment to other uses.
How do you compare two candidates with different AI job titles?
Compare the responsibilities they held on actual systems: data preparation, experimentation, integration and operations. Ask about the context in which the candidate worked and what they personally took responsibility for, then compare this experience with the remit of the role you need to fill.
Sources
- Apec : Data Engineer
- Onisep : Data scientist (expert en mégadonnées)
- Google Cloud : MLOps: Continuous delivery and automation pipelines in machine learning
- IBM : What is MLOps?
- Apec : Data analyst
- Apec : Data scientist