Role OverviewWe are seeking a
ML/RL Engineer to join our Algo team and drive the development of our unified behavioral architecture. In this role, you will help bridge the gap between simulation and the real world by developing a scalable policy framework that represents both our L4 ego-policy and a diverse population of simulated agents. You will work at the intersection of Multi-Agent Reinforcement Learning (MARL) and safety-critical system design to ensure our autonomous semi-trucks navigate highways with superhuman safety and precision.
Key Responsibilities- Behavioral Modeling: Develop and train diverse, conditioned policies that simulate realistic driving behaviors to stress-test and validate our autonomous driving stack.
- Safety-Constrained Learning: Lead the research and implementation of advanced RL algorithms to ensure safety metrics are treated as primary constraints in the learning process.
- Reward & Objective Design: Collaborate with cross-functional teams to design robust reward functions and evaluation metrics that balance safety, progress, and comfort.
- Scalable Training Pipelines: Contribute to the optimization of our large-scale, high-throughput training environments to enable rapid iteration on complex multi-agent scenarios.
- Model Architecture: Advance our state-of-the-art neural architectures to improve spatial reasoning, long-horizon planning, and interaction modeling.
- Cross-Team Collaboration: Work closely with Simulation and Planning teams to integrate research-grade models into production-quality, safety-critical software.
Required Qualifications- Professional RL Experience: Proven track record of training and deploying deep RL algorithms (e.g., PPO, SAC) for complex, real-world robotic or autonomous systems.
- Technical Mastery: Expertise in Python and PyTorch; strong understanding of modern deep learning architectures and optimization techniques.
- Academic Background: MS or PhD in Computer Science, Robotics, or a related quantitative field.
- Scientific Intuition: Ability to diagnose and solve fundamental challenges in RL training, such as variance management and distribution shift.
Preferred Qualifications- Safe RL Specialization: Experience with constrained optimization or safety-critical learning frameworks.
- Multi-Agent Systems: Background in MARL training stability, including self-play and decentralized execution strategies.
- Autonomous Driving Domain: Familiarity with vehicle dynamics and behavior planning, particularly for long-haul highway environments.
Additional Information- Compensation: Competitive salary based on experience, with opportunities for performance bonuses and equity.
- Benefits: Comprehensive health insurance, paid time off, and the opportunity to work at the forefront of the autonomous trucking industry.
