1 day ago
Sunnyvale, CA, USAStaff+
Base Salary
$336k - $370k/yr
Responsibilities
- Set the technical strategy and roadmap for the emergency trajectory model, including behavioral scope, operating envelope, system interfaces, and acceptance criteria.
- Design and train trajectory-generating policies using behavior cloning, reinforcement learning, or related sequential decision-making methods.
- Develop a rare-emergency data strategy using fleet data, targeted mining, simulation, augmentation, and reweighting.
- Create open-loop and closed-loop evaluations for collision avoidance, evasive steering, emergency braking, recovery, robustness, latency, and nominal-driving regressions.
- Lead integration into the shared driving stack and align decisions across simulation, evaluation, safety, and product engineering teams.
- Provide architecture reviews, mentoring, and clear communication of risks, trade-offs, and supporting evidence.
Requirements
- Demonstrated staff-level technical leadership setting direction for ambiguous machine learning programs, aligning multiple teams, and taking work from research through production deployment.
- Deep expertise developing learned trajectory-generation or policy models, including architecture design, objective design, training, and empirical validation.
- Hands-on experience with behavior cloning, reinforcement learning, distribution shift, robustness, and closed-loop failure analysis.
- Strong machine learning engineering skills in Python and PyTorch, including reproducible training and evaluation systems for large, heterogeneous datasets.
- Exceptional technical judgment and communication, including the ability to make safety-relevant trade-offs explicit and lead without formal authority.
- Experience applying learned models in autonomous driving or robotics and understanding motion planning, vehicle dynamics, control, or collision avoidance is desirable.
- Experience with specialist, fallback, redundant, mixture-of-experts, or model-routing architectures is desirable.
- Experience mining, generating, or evaluating rare events with simulation and fleet or real-world data is desirable.
- Experience deploying learned policies under real-time latency, reliability, and compute constraints is desirable, with proficiency in C++, CUDA, or systems optimization also desirable.
- Experience with multimodal, transformer-based, diffusion-based, or other generative trajectory or policy models is desirable.
Benefits
- Full-time role based in the Sunnyvale office.
- Hybrid working policy combining time in the office and workshops with time working from home.
- Competitive equity package.
- Inclusive interview experience with accommodations or adjustments available upon request.
Categories
ML EngineeringRobotics
