锦鲤求职岗位描述
We are seeking a skilled AI Researcher to advance foundation models for robotics, Physical AI, and agentic AI, with a focus on VLA, VLM, and World Models. This role will work on training, fine-tuning, evaluating, and optimizing large AI models for real-world robotic and edge AI applications.
The successful candidate will collaborate closely with university labs, product teams, and hardware/software platform teams to explore next-generation AI model capabilities and translate research outcomes into robust production-ready solutions.
Key Responsibilities:
Conduct research and engineering work on model training, fine-tuning, post-training, distillation, alignment, and simulation-based optimization for robotics and agentic AI scenarios.
Research and prototype agentic AI systems that combine foundation models with planning, tool use, memory, retrieval, environment interaction, and task execution.
Improve model capabilities for agentic workflows, including function calling, tool selection, multi-step reasoning, self-reflection, task decomposition, long-horizon planning, and failure recovery.
Collaborate closely with university labs and research partners to explore emerging model architectures, evaluate research prototypes, and translate promising technologies into deployable solutions.
Required Qualifications:
BS/MS degree in Computer Science, Software Engineering, Electrical Engineering, Robotics, Artificial Intelligence, or equivalent practical experience.
Solid understanding of the architecture, training, fine-tuning, and deployment methodologies of mainstream large AI models, including LLMs, VLMs, VLA models, or World Models, with an ability to actively track research progress and assess practical value.
Familiarity with agentic AI concepts such as tool use, function calling, planning, memory, RAG, autonomous task execution, and multi-step reasoning.
Understanding of agent system design patterns, including planner-executor, ReAct-style reasoning, retrieval-augmented agents, tool-using agents, and multi-agent collaboration.
Excellent communication skills in English, with the ability to collaborate effectively across research, engineering, product, and external academic teams.
Strong programming skills in Python and Shell; familiarity with C/C++ for model integration, deployment, or performance optimization.
Preferred Qualifications:
Experience with supervised fine-tuning, LoRA/QLoRA, knowledge distillation, preference optimization, continual learning, imitation learning, reinforcement learning, and simulation-to-real transfer.
Hands-on experience with open-source robotics or Physical AI models such as OpenVLA, RT-1/RT-2, RDT, diffusion policy models, pi0-style models, or related architectures.
Experience in building tool-using agents, local assistants, robotics agents, workflow agents, code agents, or domain-specific autonomous systems.
Experience designing agent evaluation methods, including task success rate, tool-call accuracy, planning quality, grounding quality, safety, robustness, and latency.