Senior Reinforcement Learning Control Engineer

All3 is transforming how buildings are conceived, developed and delivered. We combine AI-powered design with robotic prefabrication and on-site assembly to build custom architecture at the cost and speed of mass production - unlocking even the most complex sites. We’re currently seeking a Senior/Lead Reinforcement Learning Control Engineer (seniority level adjustable based on candidate expertise) to develop learning-based control systems for dynamic locomanipulation on our large legged robotic platform, Mantis. You’ll design reinforcement learning solutions that integrate tightly with the rest of our framework to enable robust locomotion and manipulation in complex, semi-structured construction environments.

Responsibilities:

• Design and train reinforcement learning policies using physically grounded sensor models (LiDAR, cameras, IMUs) to support autonomous locomotion and manipulation capabilities.
• Integrate learned components within a complex software architecture, ensuring stability and reliability.
• Develop a robust methodology for policy training, tracking improvements and monitor progress and performance evaluation.
• Develop robust sim-to-real transfer strategies and ensure successful deployment of the policies to the hardware platform.
• Analyse hardware experiments, identify failure modes and iteratively improve performance and robustness.
• Collaborate closely with perception, state estimation and software engineers to deliver cohesive, real-time robotic behaviour and evolve the robot software architecture design.
• Contribute to the evolution of a unified whole-body control framework for loco-manipulation capabilities.

This is not a simulation-only research role. The systems you build will run on hardware and must operate reliably under real-world constraints.

Expertise:

• Strong proficiency in Python and PyTorch; with a good C++ experience.
• Solid understanding of ML approaches and their challenges and practical limitations in physical systems.
• Hands-on experience with legged locomotion (quadruped or biped) and/or robotic manipulation.
• Good understanding of robot dynamics, kinematics and contact modelling.
• Experience deploying learning-based components on real robotic hardware.
• Ability to reason about stability, safety and robustness — not just reward optimisation.
• Ability to shape the technical direction by helping design methodology, provide in-depth analysis and insightful feedback.

Nice to have:

• Experience with whole-body control of legged systems.
• Experience combining classical control and learning in hybrid architectures.
• Exposure to state estimation and multi-sensor integration.
• Experience with contact-rich manipulation.
• Experience in /Exposure to construction robotics, heavy machinery or large-scale physical systems.
• Experience with safety certification of complex robotic systems operating in contact rich environments.

We offer:

• Chance to be a part of a large-scale project.
• Team driven by impactful cause.
• Hybrid format of work with the lab located in Park Royal.
• Private dental or full medical (dental treatments aren’t covered) insurance.
• Flexible working schedule.
• 28 days of annual leave.

At All3, you’ll work on one of the hardest problems in robotics: combining dynamic locomotion and manipulation into a structured, deployable autonomy system. You’ll join a small, focused team building real machines for real environments - where physics matters, reliability matters and thoughtful engineering wins over hype.