AI Hiring
Which AI discipline does the work require?
Titles overlap across organisations. Start with the decisions and deliverables a person will own, then use the title to describe that work clearly.
At a glance
Three disciplines. Different responsibilities.
AI engineer
Focus: the application. Typical output: a model-powered feature or workflow.
ML engineer
Focus: the model lifecycle. Typical output: a repeatable training, deployment or monitoring system.
Data scientist
Focus: the analytical question. Typical output: an experiment, model or analysis that informs a decision.
These are common patterns, not fixed boundaries. Define what the person will own before choosing the title.
In this article
AI engineering
An applied AI role may focus on integrating models into a product or workflow, evaluating outputs and connecting the application with existing systems. For a generative AI application, ask about evaluation design, retrieval quality, error handling and how the person worked with users and product owners.
ML engineering
An ML engineering requirement may centre on the systems around model training, deployment and operation. Clarify responsibility for pipelines, reproducibility, serving, monitoring and changes over time. A research prototype and a maintained service have different operational requirements.
Data science
A data science role may focus on framing analytical questions, examining data, experimentation and modelling. Establish whether the team needs causal reasoning, forecasting, statistical analysis or another specific discipline. Ask how findings informed a decision and how limitations were explained.
Make the boundaries explicit
List the outputs, interfaces and decisions each role owns. Identify who is responsible for data quality, software integration and product priorities. These responsibilities can sit in different roles; the aim is clear ownership rather than a universal organisation chart.
A useful interview prompt
Ask a specialist to explain one relevant project from problem definition to practical use: their personal contribution, a decision they changed, the evidence behind it and what they would do differently. Follow the details rather than scoring the number of tools mentioned.
For a contract requirement, start with the gap in your existing team. A model already developed by a data scientist may need an ML engineer to make deployment repeatable, or an AI engineer to connect it to a product. Explain what is already working, what needs to change and who will maintain it after the engagement. Avoid expecting one specialist to cover research, application development and platform operation without appropriate support.
Put the requirement into practice
Build the capability you need.
Discuss your priorities, the experience you need and the team around the role with Theseus.
Discuss your requirement
