Senior Engineer; m​/f​/d

Role Overview

This role focuses on defining and advancing learning-based approaches for motion planning in autonomous systems. It involves setting the technical direction and delivering solutions that evolve from traditional rule-based and optimization-driven methods toward data-driven techniques such as imitation learning, reinforcement learning, and diffusion-based planning.

The position combines hands‑on technical leadership with ownership of the roadmap, ensuring that learning-based planners are not only innovative but also practical, reliable, and ready for real-world deployment. The emphasis is on delivering measurable improvements over existing planning approaches through robust engineering and evaluation.

• Technical roadmap ownership: Define and execute a clear roadmap for learning-based planning over a 12–24 month horizon. Identify the most suitable methods, determine build vs. adopt decisions, and establish criteria for progressing solutions from prototype to production deployment.
• Model development and training: Design, train, and refine learning-based planning models using a range of approaches, including imitation learning, reinforcement learning (online and offline), diffusion models, and hybrid techniques. Select the most effective method based on the problem context.
• System integration: Integrate learning-based planners into existing planning and control architectures. Ensure clean interfaces with upstream perception systems and downstream control layers, along with reliable fallback mechanisms and production-grade robustness.
• Data strategy and requirements: Define data needs for training and evaluation, including sources such as human demonstrations, simulation, and existing planning outputs. Drive efforts to ensure sufficient coverage of scenarios, edge cases, and real-world variability.
• Evaluation and benchmarking: Develop comprehensive evaluation frameworks to assess learned planners against established baselines. Establish metrics, scenario-based testing, and failure analysis processes to ensure readiness for deployment.
• Iterative experimentation: Lead the full experimentation loop—from hypothesis and model development to real-world testing and performance analysis—ensuring continuous improvement based on empirical results.
• Hybrid planning strategy: Work closely with existing planning approaches to determine where learning-based methods provide value and where traditional techniques remain optimal. Contribute to designing hybrid systems that combine both effectively.

• 6+ years of experience in motion planning for autonomous systems or mobile robotics, including hands‑on work with learning-based planning approaches.
• Demonstrated experience taking learning-based planning solutions from concept through to deployment in real-world environments.
• Strong understanding of classical planning approaches, including model predictive control (MPC), sampling-based, and optimization-based methods.
• Proficiency in C++ and Python, with experience in ROS (ROS1 and/or ROS2).
• Experience with modern machine learning frameworks (e.g., PyTorch) and solid understanding of experimental design and validation.

• Experience publishing or delivering solutions involving learning-based planning methods such as imitation learning, reinforcement learning, diffusion-based models, or world-model approaches.
• Background in designing hybrid systems that combine learned components with classical planning or safety mechanisms.
• Familiarity with building robust, real‑world systems that operate under uncertainty and dynamic conditions.