How the robotics engineer skill set is changing in the age of physical AI

# Overview Jakub Tomaszewski, a senior robotics engineer, traces how the role of a robotics engineer has changed since 2020. Faster tools, powerful simulation platforms, hardware acceleration, and AI have reduced time spent on low-level implementation. That lowers the barrier for writing code but raises the bar for deciding how systems should be designed, integrated, and validated.

# Four expanded competency areas

Physical AI and reinforcement learning Learning-based methods are no longer just research topics. Reinforcement learning can improve controller robustness in stochastic or changing conditions. Physical AI extends AI beyond perception and planning to systems that perceive, act, and interact with the physical world. Senior engineers must understand how to mix learning-based methods with classical control, rather than treating them as separate paths.

Interdisciplinary knowledge and system integration Robotics increasingly requires tight integration across mechanics, electronics, control, perception, software architecture, and validation. Senior, lead, and architect roles have shifted toward system architecture and integration. Building a digital twin, for example, requires a 3D model, physics description, sensor configuration, and telemetry paths linking simulator and physical machine.

# How to learn effectively as an experienced engineer

# What foundational skills still matter Mathematics and robotics theory remain central. Tomaszewski argues that fundamentals—kinematics, dynamics, estimation, control, and perception—are the building blocks of system architecture. Even with tools that automate implementation, an engineer must be able to judge whether a model is well posed, a controller is stable, or a sim-to-real gap is acceptable. Without those abilities, architecture becomes an ungrounded design.

# Practical implications for senior engineers

• Allocate more time to architecture, integration, and validation rather than low-level implementation.
• Invest in simulation and digital twin capabilities to test designs before hardware commits.
• Learn when and how to combine reinforcement learning and other learning-based methods with classical control approaches.
• Preserve and deepen mathematical and theoretical foundations to assess models and safety margins.

# Bottom line Tools have accelerated many implementation tasks, but senior robotics engineers must absorb broader responsibilities: design complete systems, integrate diverse components, validate behavior in simulated and real environments, and retain the theoretical grounding needed to judge correctness and safety.