Robotics Software Engineer – AI Systems

About RoboTensor

RoboTensor is a Physical AI research and development company building the infrastructure that turns robot foundation models into specialized physical capabilities.

The real world is the final benchmark. That means learned policies ultimately need to operate reliably through real cameras, sensors, compute systems, control interfaces, and machines.

About the Role

We are looking for a Robotics Software Engineer – AI Systems to build the software layer connecting learned intelligence to physical machines.

You will own the systems required to run, evaluate, observe, and improve AI policies on real robots.

You will work at the boundary between machine learning and physical systems—turning model outputs into reliable robot behavior and turning robot experience back into useful data for the next training cycle.

What You’ll Do

• Build the software infrastructure required to deploy learned policies on physical systems.
• Integrate cameras, sensors, actuators, grippers, controllers, and robot interfaces.
• Build reliable interfaces between AI models and robot control systems.
• Develop real-time and near-real-time inference pipelines.
• Build systems for collecting synchronized observations, actions, video, state, and telemetry from physical runs.
• Develop tooling for running experiments and evaluations repeatedly on real machines.
• Build logging, monitoring, replay, and debugging infrastructure for physical AI systems.
• Diagnose failures across model inference, software, networking, sensors, control interfaces, and hardware.
• Improve system reliability, latency, recoverability, and observability.
• Build interfaces that allow researchers to deploy new policies safely and quickly.
• Connect real-world evaluation results back into training and data pipelines.
• Help standardize how RoboTensor interfaces with different robot embodiments and hardware platforms.

What We’re Looking For

• Strong software engineering fundamentals.
• Experience building reliable software that interacts with physical or real-time systems.
• Strong Python and/or C++ skills.
• Experience debugging complex systems across multiple layers.
• Familiarity with networking, concurrency, processes, hardware interfaces, and system performance.
• Ability to design clean abstractions around messy hardware and vendor-specific interfaces.
• Comfortable working closely with ML engineers and understanding the needs of learned policies.
• Strong ownership mentality around reliability: if an experiment fails for a systems reason, you want to understand and eliminate it.

Nice to Have

• Experience with robotics software or autonomous systems.
• Experience with ROS/ROS 2 or similar middleware.
• Experience integrating industrial arms, mobile robots, grippers, cameras, or other robotic hardware.
• Experience deploying GPU-based inference systems.
• Experience with real-time systems and control loops.
• Experience with calibration, camera systems, or state estimation.
• Experience building automated test and evaluation infrastructure for physical systems.
• Familiarity with VLA models or learned robot policies.

What Success Looks Like

A researcher can take a newly trained policy and run it on a real system without rebuilding the software stack around it.

Experiments are repeatable. Failures are observable. Data is captured correctly. New robots can be integrated efficiently.

The physical system becomes part of the learning loop rather than a fragile final demonstration.