Senior Reinforcement Learning Engineer
ANYbotics · Zurich, Switzerland · Posted May 26, 2026
Transparency
DeadOnline for 25 weeks and never reposted. Jobs this old or reposted this often rarely lead to a hire. It may be filled already.
- First seen
- Oct 6, 2026
- Offered for
- 178 days
Overview
- Category
- AI & machine learning
- Employment type
- Permanent
- Workplace type
- Hybrid
- Country
- Switzerland
- Seniority
- Senior
- Schedule
- Full time
Key skills
- Reinforcement Learning
- Design, train and optimise RL policies for robot motion from simulation to field.
- Sim‑to‑Real Transfer
- Build pipelines that ensure policies trained in Gazebo or Isaac Sim work reliably on hardware.
- Python and C++ Development
- Write, deploy and maintain clean, high‑performance code for the learning and locomotion stack.
- Robotic Simulation
- Use Gazebo or Isaac Sim to prototype and evaluate motion control, state estimation and planning.
- Linux and Middleware Integration
- Integrate learned components into the robot software stack using Linux and middleware frameworks.
Key phrases
- reinforcement learning
- policy optimisation
- sim-to-real policy transfer
- reward engineering
- mission‑critical, production deployments
- Gazebo or Isaac Sim
- Python and C++ software
- PyTorch
- Linux systems
- middleware frameworks
- motion control
- state estimation
- path planning
- domain randomisation
- policy robustness
From the posting
The Opportunity
We are looking for a Senior Reinforcement Learning Engineer to join our Motion Intelligence team. You should bring deep expertise in reinforcement learning and policy optimisation, grounded in real-world robotics experience, and be excited to push the boundaries of autonomous robots. We are looking for someone who can provide senior-level RL and learning-based control guidance across the team — shaping how we train, deploy, and maintain learned policies on physical robots operating in demanding industrial environments.
Beyond RL expertise, we value engineers who take pride in writing clean, maintainable code, simplifying complex systems through thoughtful design, and championing sound interfaces and architecture.
At ANYbotics, we develop tailored software solutions that enable our legged robots to execute inspection and maintenance tasks with precision and consistency. This role sits at the intersection of cutting-edge RL research and production robotics — from sim-to-real policy transfer and reward engineering through to reliable deployment on mission-critical systems in the field. You’ll work across the full lifecycle — from early development prototyping through to mission‑critical, production deployments.
The Market & Our Technology
ANYbotics transforms industrial plants in the energy, process, and utility sector by introducing robotics to a wide range of novel applications that so far were beyond reach. Our customers are large asset operators and industrial service providers pioneering the use of robotics technology for inspection and maintenance. Our mobile robot ANYmal uses legs for extreme mobility in complex environments, camera- and LIDAR-based sensing for full autonomy and obstacle avoidance, and AI for high-quality and consistent inspection results. We develop numerous customised hardware systems, including the entire robotic platform, actuators, sensors, inspection payloads, charging systems, and all related ANYbotics electrical hardware.
Your Impact
- Lead the design, training, and deployment of reinforcement learning policies for robot motion — bridging the gap from simulation to reliable real-world performance
- Provide senior technical guidance on RL and learning-based control across the team, mentoring engineers and establishing best practices for policy development workflows
- Own and evolve the RL training infrastructure and sim-to-real pipeline, ensuring reproducibility, scalability, and fast iteration cycles
- Shape the technical vision for internal ML tooling and experiment management (e.g. training dashboards, automated evaluation pipelines), driving efficiency and rigour across the team's learning workflows
- Collaborate closely with cross-functional stakeholders to identify how to expand the robot's autonomous operational envelope
- Triage field issues related to locomotion, recognise failure patterns, and rapidly improve policy robustness based on real deployment data
- Write, deploy, and maintain efficient Python and C++ software for the learning and locomotion stack
Your Profile
- PhD in robotics, machine learning, computer science or a related field with a strong focus on reinforcement learning; alternatively, an equivalent track record of RL research and deployment in robotics Or
- Master's degree from a top-tier technical university (e.g. ETH Zurich, EPFL) in robotics, machine learning, computer science or related field and 5+ years of professional experience
- Proven track record of shipping ML models to the field and maintaining those solutions over time
- Solid grounding in robot control fundamentals and autonomous systems, including: motion control, state estimation, path planning and actuation
- Experience using robotic simulation tools such as Gazebo or Isaac Sim
- Strong understanding of sim-to-real transfer, domain randomisation, reward shaping, and policy robustness techniques
- Proficiency in Python and modern ML frameworks (PyTorch); working knowledge of C++
- Strong knowledge of Linux systems and middleware frameworks for integrating learned components into a larger software stack
- Pragmatic and solution-oriented mindset — comfortable balancing research exploration with production delivery
- Excellent communication skills in English
Bonus Points
- Experience training and deploying RL policies on physical robots — not just in simulation
- Development of scalable and modular robot architectures of motion control systems
- Experience with navigation systems and autonomous mobile robot operation in unstructured environments
- Interest in agentic engineering toolchains
- Experience leading software architecture design and software engineering best practices
- A solid grasp of physical systems, specifically multibody dynamics, electromechanical drive physics, energy optimization, and contact physics, to ensure learned policies respect hardware limits and transfer successfully from simulation to the real world
