I get this question from two sides. People who hire ask me what to look for. Students ask me what to study. It is the same question, so here is one answer for both.
A warning first: there is no such thing as a standard Physical AI engineer yet. The field is being put together from three older ones, and most good people I have worked with came in through one door and learned the other two rooms on the job.
The short answer
A Physical AI engineer builds the intelligence of machines that sense and act in the real world. The job sits where machine learning, robotics and embedded systems meet, and the rare skill is being comfortable in all three.
Three doors into the field
| Coming from | What they bring | What they usually have to learn |
|---|---|---|
| Machine learning | Models, training, data | That the machine has 5 watts, not 500, and that a late answer is a wrong answer |
| Robotics and control | Dynamics, feedback, safety, respect for hardware | Learning-based methods, and living with a system you cannot fully write down |
| Embedded and hardware | Power, memory, timing, sensors | What the models need, and how to give it to them cheaply |
Seven skills that matter
- Perception. Turning cameras, touch, sound and motion sensors into something a machine can act on.
- Control and timing. Closing the loop between sensing and acting, in milliseconds, every time.
- Working inside a budget. Power, memory and compute are part of the problem, not an afterthought.
- Sensors and signals. Knowing what a sensor really measures, and what noise looks like.
- Simulation. Using it for what it is good at, and knowing where it lies.
- Data from the real world. Collecting it, cleaning it, and noticing when it has drifted.
- Testing on hardware. Patience, method and a healthy fear of things that move.
What I look for beyond the list
- They have built something that moved. And it broke, and they can tell you why.
- They ask about constraints first. How much power? How fast? What happens if it is wrong?
- They can talk to the other two rooms. A learning person who can read a datasheet, a hardware person who can read a loss curve.
- They are curious about biology. Not required… but animals solved most of these problems already, and it shows in how people think.
If you are a student
Build a small machine that senses and acts. A cheap microcontroller, a camera, two motors. Make it follow something, then make it learn something, then run it on a battery. You will meet latency, noise and power in the first week, and you will understand this page better than by reading it.
Then read some of the old papers. I keep a short list here: a Physical AI reading list.
Job titles you will see
The title “Physical AI engineer” is still rare. The same work is advertised as robotics machine learning engineer, robot learning researcher, perception engineer, controls engineer, embedded machine learning engineer or edge AI engineer. Read the description, not the title.
Questions I get
What does a Physical AI engineer do?
A Physical AI engineer builds the intelligence of machines that sense and act in the real world: robots, drones, vehicles, industrial equipment. The work sits between machine learning, robotics and embedded systems.
What skills does a Physical AI engineer need?
Machine learning for perception and control, real-time and control engineering, embedded computing under tight power and memory budgets, sensors and signal processing, simulation, and disciplined testing on real hardware.
What should I study to work in Physical AI?
There is no single degree. People come from robotics, computer science, electrical engineering, physics and computational neuroscience. What they share is that they have built something that had to work outside a computer.