There is a quiet assumption that AI belongs to the computing branches. It does not. Some of the most valuable applications sit in civil, mechanical and electrical engineering, precisely because the people who understand those domains are not usually the people building models.

Where AI sits in an engineering workflowJudgement — what problem is worth solvingDomain knowledge — is this answer physically sane?AI assistance — drafts, code, summaries, first passesTooling — CAD, compilers, simulators, instrumentsAI moves the floor up. It does not move the ceiling down.
Some of the most useful applications of machine learning are in branches that do not teach it as a core subject.

Where it actually applies

  • Civil. Structural health monitoring from sensor data, predicting concrete strength from mix and curing variables, analysing traffic and transport data, detecting defects from site imagery.
  • Mechanical. Predictive maintenance from vibration and temperature data, optimising process parameters, surrogate models that approximate expensive simulations, quality inspection from images.
  • Electrical. Load forecasting, fault classification on distribution networks, state-of-health estimation for batteries, renewable generation forecasting.
  • Automobile. Engine and emissions calibration from test-bench data, driver behaviour analysis, battery thermal management.

What you actually need to learn

Less than you think, and in a specific order. Python for handling data, the basics of statistics, and one practical machine learning library. What you do not need is the full computer science curriculum, because you are not building the tooling, you are applying it to a domain you already understand.

Your advantage over a computing student is exactly that domain understanding. You can tell when a model's output is physically impossible. That is a judgement they cannot easily acquire.

How to start without derailing your degree

Use your own coursework as the dataset. Laboratory sessions in every core branch generate real measurements. Analysing your own laboratory data with a simple model teaches both the method and its limits, and it costs you no extra subject.

Then aim a mini project or final-year project at the boundary. A project that applies a data method to a genuine problem in your discipline is more distinctive at interview than either half would be alone.

What to take away

  • The highest-value AI applications often sit in core branches, not computing branches.
  • You need Python, statistics and one library, not a second degree.
  • Domain knowledge is your advantage: you can tell when an output is physically impossible.
  • Use your own laboratory data, then aim a project at the boundary.

Thinking about engineering at Maria?

Our admission office can talk you through branches, eligibility and the counselling process for 2026–27. No obligation.