The posting
Vision-Language-Action(VLA) 및 imitation learning·강화학습 기반의 manipulation policy를 개발하여, 실제 로봇의 task 성공률과 일반화(generalization) 성능을 지속적으로 끌어올리는 역할입니다. 다양한 작업과 환경에서 로봇이 스스로 잘 동작하도록 만드는, AI Core의 핵심 학습 축을 담당합니다.
In this role you build manipulation policies based on Vision-Language-Action (VLA), imitation learning, and reinforcement learning — continuously improving real-robot task success and generalization. You own a core learning axis of AI Core: making robots perform reliably across diverse tasks and environments.
주요업무 (Key Responsibility)
- VLA·imitation learning·RL 기반 manipulation policy 개발
- 실 로봇의 task success rate 및 generalization 성능 측정·개선
- 대규모 demonstration 데이터를 활용한 policy 학습 파이프라인 구축
- 다양한 task와 환경 일반화를 위한 모델 구조·학습 전략 실험
- AI Core·데이터·로봇 OS 팀과 협업하여 학습된 policy를 실 배포로 연결
- Develop manipulation policies based on VLA, imitation learning, and RL
- Measure and improve real-robot task success rate and generalization
- Build policy training pipelines that leverage large-scale demonstration data
- Experiment with model architectures and training strategies for cross-task/environment generalization
- Collaborate with AI Core, Data, and Robot OS teams to move trained policies into deployment
Requirements
- Robot learning, manipulation policy, 또는 VLA 관련 3~5년의 연구·개발 경험
- Imitation learning, reinforcement learning, 또는 foundation model 기반 policy 학습 경험
- Python·PyTorch 기반 대규모 모델 학습 및 실험 역량
- 실 로봇 또는 시뮬레이터에서 policy를 학습·평가해본 경험
- 컴퓨터공학·AI·로봇공학 관련 석사 이상 또는 그에 준하는 경험
- 3–5 years of research/engineering experience in robot learning, manipulation policy, or VLA
- Experience with imitation learning, reinforcement learning, or foundation-model-based policy learning
- Strong large-scale model training skills in Python and PyTorch
- Experience training and evaluating policies on real robots or simulators
- Master's degree in CS, AI, Robotics, or equivalent experience
Preferred
- VLA, robot foundation model(π0, NVIDIA Gr00t 등), large-scale imitation learning 관련 연구 실적
- 대규모 로봇 demonstration 데이터셋 구축·활용 경험
- Sim-to-real 및 multi-task generalization 경험
- 로봇공학·AI 석/박사 학위
- 실제 제품·현장에 학습 policy를 배포해본 경험
- Research track record in VLA, robot foundation models (π0, NVIDIA Gr00t, etc.), or large-scale imitation learning
- Experience building/using large robot demonstration datasets
- Sim-to-real and multi-task generalization experience
- Master/PhD in Robotics or AI
- Experience deploying learned policies into products or the field
Benefits
- Claude & ChatGPT subscriptions provided (Claude · ChatGPT 유료 플랜 지원)
- Minimal meetings with fast decision-making (불필요한 회의를 최소화하고 빠르게 의사결정합니다.)
- Modern intranet/tools (Google Workspace, Slack, Notion, Linear, Workable, Flex.team 등)
