ML Engineer vs AI Engineer
Two job adverts sit side by side: "ML Engineer" and "AI Engineer". Both mention models, Python and deployment, and it is easy to think they are the same job with different titles. They overlap, but their day-to-day work starts from opposite ends.
The rule
ML engineers build models from data; AI engineers build products from models.
An ML engineer starts with the organisation's own data. At the letting agency, that means thousands of past lettings like our five flats. The work is choosing features (size, rooms, metro distance), training a model, measuring its error, and running it in production. The model's weights, the numbers it learned, are made in-house.
An AI engineer starts with a model someone else has already trained, usually a large pretrained model reached through an API. The agency might want an assistant that answers tenants' questions about their lease. Nobody at the agency trains the language model. The work is:
- writing and testing prompts,
- retrieving the right lease clauses and putting them in front of the model,
- controlling cost and speed per request,
- adding guardrails so the assistant does not invent rules or leak data,
- evaluating answers so quality can be measured, not guessed.
Both roles share a lot: honest evaluation, deployment, monitoring and good software habits. Job titles also vary between companies, so read the duties, not just the title.
The shortcut
Ask one question: "Where do the model's weights come from?"
- The team trains them on its own data → ML engineering.
- They arrive ready-made and the team shapes what goes in and comes out → AI engineering.
- The team takes ready-made weights and trains them further on its own data (fine-tuning) → the borderline where the two meet.
Rent predictor
Trained from scratch on 50,000 past lettings. The weights are made here, so this is ML engineering.
The trap is to sort by the words in the project name. "AI-powered rent estimate" is still ML engineering if the team trains the model on its own lettings.
Drills
Answer fast — trust the rule.
A team collects 50,000 past lettings and trains a model to predict rent. Which role leads this?
A team connects a hosted language model to the agency's lease documents so tenants can ask questions. Which role leads this?
Which skill do both roles need most?
What is the quickest question for telling the two roles apart on a project?
Remember
- ML engineers build models from data; AI engineers build products from models.
- Shortcut: ask where the weights come from. Fine-tuning sits on the border.
- Trap: sorting by buzzwords or titles instead of the actual work.