Careers and industry
How robotics work is organised and how to get into it. Industrial, service and research robotics, the main engineering roles and the skills behind them, interviews, portfolios and the market numbers.
Overview
Robotics is not one job. A working robot needs mechanical and electrical design, motor control, embedded software, perception, planning, integration on site, testing, safety engineering and support. Most engineers specialise in one or two of these and know enough of the rest to work across the boundaries. This hub is for students and for engineers moving into robotics from software, controls, mechanical or machine learning work. It describes how the industry is split, which roles exist, what interviews test, and how to build a portfolio that shows what you can do.
Key ideas
Industrial, service and research robotics
Industrial robotics means arms in factories: welding, painting, assembly, machine tending and palletising. Customers often buy a complete cell from a system integrator, and the work combines robot programming in the maker’s own environment, PLCs, safety design and long commissioning trips. According to the International Federation of Robotics (IFR), factories worldwide installed more than 600,000 industrial robots in 2025, an 11% rise, and the operational stock reached a record 5 million units. The IFR forecasts 655,000 installations in 2026 and 806,000 in 2029. It reports that China installed 354,000 units in 2025, 59% of the world total.
Service robotics covers robots outside the factory cell: mobile robots in warehouses and hospitals, cleaning, delivery, inspection, agriculture and drones. The work leans towards autonomy software, fleet operations and field reliability, and products change faster. Research robotics, in universities and company labs, works on problems that are not yet products, such as learned manipulation, legged locomotion and humanoids. Its output is papers, open-source code and prototypes.
Roles and the skills behind them
A common core runs through nearly every robotics role. It includes frames and transforms, kinematics, feedback control and state estimation, together with Python and C++, Linux, Git and ROS 2. On top of that core sits a specialism. Controls engineers tune loops and model dynamics. Perception engineers turn camera and lidar data into poses and maps. Planning engineers make motion fast and collision-free. Integration and field engineers make a cell or fleet work at a customer site, which is where a great deal of practical robotics knowledge is earned. Safety engineers own the risk assessment and the safety functions, see the safety hub. Robot learning roles add data collection and training, with machine learning background from the AI/ML Field Guide.
Interviews and portfolios
Prepare to work through fundamentals on a whiteboard or in a shared editor. Compose two transforms and explain the frame each one maps between. Derive the forward kinematics of a two-link arm. Explain how you would tune a PID loop on a real joint and what you would log. Explain what a Kalman filter assumes. Debug a robot that drifts or oscillates from a short description. Practise coding in both Python and C++. A take-home task may ask you to write a small ROS 2 node or planner.
A portfolio should show one or two finished projects rather than many started ones. Each needs a README that says what the system does, a short video, the numbers you measured, and what went wrong and how you fixed it. A simulated project is fine if it is complete and honest about its limits. A merged pull request to a widely used open-source package, such as a ROS 2 driver, MoveIt 2 or Nav2, shows that you can read other people’s code and work with reviewers.
Where to start
The beginner track runs from What is a robot to Hello ROS 2 and covers the common core above. For portfolio pieces, finish one of the projects: the two-link arm that draws with IK, the differential drive A* and Kalman project, the ROS 2 talker, listener and tf2 project or the PyBullet pick-and-place project. Then extend it in a direction that matches the role you want. The software matrix and news pages show which tools are in use now, and events lists conferences and courses.
Tools and platforms
| Tool | What to show with it | Licence |
|---|---|---|
| ROS 2 | A small system of nodes, launch files, parameters and tf2 | Apache-2.0 (core client libraries) |
| MoveIt 2 | Arm planning with collision objects and a pick task | BSD-3-Clause |
| Nav2 | A tuned navigation setup for a simulated base | Mixed by package |
| MuJoCo | A controller or policy with measured tracking error | Apache-2.0 |
| Gazebo | A full simulated robot with sensors and ROS 2 | Apache-2.0 |
| LeRobot | A recorded dataset and a trained imitation policy | Apache-2.0 |
| PX4 | A simulated flight with offboard control from ROS 2 | BSD-3-Clause |
Common mistakes
- Listing every tool you have touched. Show depth in a few, with results.
- Skipping the maths. Interviews return to frames, kinematics, control and estimation.
- Building only in simulation and never discussing what would change on hardware.
- Ignoring industrial robotics. Integration and field roles teach a great deal about real robots.
- Quoting market numbers without a source. Use the primary report and say whose figure it is.
- Treating a demo video as proof of a company’s maturity when choosing an employer. Ask about products, customers and deployments.
Going deeper
- International Federation of Robotics, Five Million Robots now Operate in Factories Globally, press release, 24 September 2026.
- Lynch and Park, Modern Robotics, Cambridge University Press, 2017, with course material from Northwestern.
- Tedrake, Robotic Manipulation and Underactuated Robotics, MIT course notes, free online.
- LaValle, Planning Algorithms, Cambridge University Press, 2006, free online.
- Macenski et al., Robot Operating System 2: Design, architecture, and uses in the wild, Science Robotics, 2022.