30 days ago
Sŏngnam, Korea, SouthMid Level
Responsibilities
- Design VLA policy and simulator observation/action interfaces and develop closed-loop reinforcement learning post-training using policy optimization methods such as GRPO and PPO.
- Build self-play training structures in vector- or occupancy-based driving simulators where the ego vehicle and traffic agents interact with current policies and historical checkpoints.
- Roll out candidate actions and trajectories across possible future scenarios and co-develop action selection based on expected return and safety constraints.
- Validate policy safety and generalization with closed-loop benchmarks and analyze reward hacking and simulator bias.
- Lead promising reinforcement learning ideas from design and large-scale training through rigorous evaluation and integration into driving models.
Requirements
- At least 3 years of experience in machine learning, reinforcement learning, or robotics, or equivalent research and development capability.
- Experience implementing policy learning algorithms and conducting training and evaluation with Python and PyTorch.
- Understanding of policy optimization, reward design, and credit assignment, with the ability to design reinforcement learning experiments.
- Experience training policies in simulation or multi-agent environments and validating improvements against baselines.
- English communication skills sufficient for discussing technical issues and collaborating with overseas offices.
- Preferred: experience developing self-play, multi-agent reinforcement learning, or policy-population training structures.
- Preferred: experience applying GRPO or PPO reinforcement learning post-training to VLA or generative policies.
- Preferred: experience training policies in vector- or occupancy-based driving simulators or developing world-model-based rollout planning.
- Preferred: experience with large-scale parallel rollout, distributed training, or sim-to-real validation.
Benefits
- The interview process consists of application review, a roughly one-hour video interview, and a second interview conducted in person or by video for approximately three hours.
- A three-month probationary period may apply.
- Veterans and employment-protected candidates, as well as registered individuals with disabilities, receive preference according to applicable law.
Categories
ML EngineeringRobotics
About 42dot
42dot builds software-defined vehicle and autonomous mobility platforms for automakers and mobility service providers. Its products include AI-driven vehicle software, data-based update services, and core platform components such as API gateways and identity/access management to connect services securely. The privately held company is headquartered in Seongnam-si, Gyeonggi-do, and focuses on transportation and automotive markets.