VLA Lead

1. Possess the fundamental ability to independently develop core algorithms for VLA (Vision-Language-Action) robot models2. Capable of integrating multimodal data into continuous action prediction models to support long-term decision-making3. Familiar with core generative model technologies such as diffusion, flow matching, and Transformers; able to adapt or innovate these to support robot action learning; capable of validating model training results on small robots or in simulation environments4. Possess original algorithmic experience in the field of large-scale robot models and be capable of designing architectures from scratch.5. Be familiar with physical world modeling and Sim-to-Real pipelines, and be able to directly implement algorithms into robotic actions.6. Have experience conducting research on large-scale VLA, multimodal embodied models, or world models at top-tier laboratories or companies, with relevant publications or open-source project experience.