What Is an AI Engineer?

Pedal Point is a small online bike shop drowning in the same support questions: "Where is my order?", "Can I return this helmet?", "Will this tyre fit my wheel?". A few years ago, automating those answers meant hiring a research team. Today one engineer can build a helpful assistant in weeks, by standing on a model somebody else trained. That engineer is an AI engineer, and this course follows them as they build the Pedal Point assistant.

The rule

AI engineers build on models; ML engineers build the models.

A machine learning (ML) engineer gathers data, designs and trains models, and tunes them so they learn well. Their main output is a model: a large file of numbers called weights.

An AI engineer usually starts with a model that already exists, often a large language model offered through an API or downloaded as open weights. Their output is a product. The daily work looks like this:

• choosing a model that fits the task, the budget and the speed needed
• writing the instructions the model follows
• feeding it the right data, such as policies or a customer's order
• letting it call tools, like "look up order status"
• testing its answers, watching cost and speed, and keeping it safe

The two roles overlap, and many people do both. But the centre of gravity differs: one shapes the model, the other shapes everything around it.

This shift changed product development. A prototype that talks can exist in an afternoon, so ideas get tested sooner. The hard part moved from "can we build a model?" to "can we make it reliable?", because a model can give different answers to the same question and can sound confident while wrong.

A word on hype: AGI, artificial general intelligence, names a hoped-for system that could match people across almost any intellectual task. The Pedal Point assistant is not that. It is a capable tool pointed at one job, and treating it that way keeps your designs honest.

The shortcut

Ask one question of any task: "Does this change the model's weights?" If yes, it leans toward ML engineering. If the weights stay fixed and you are building around them, it is AI engineering.

The trap is thinking an AI engineer's job ends when the demo works. The demo is the easy half; reliability is the job.

Drills

Answer fast — trust the rule.

Remember

• AI engineers build on models; ML engineers build the models.
• The test: does this work change the weights?
• A working demo is the start. Reliability, cost and safety are the job.