researcher（pre-training）

Responsibilities

• Drive technical initiatives in Vision-Language-Action (VLA) model development through innovative pre-training methodologies (e.g., cross-modal curriculum learning).
• Monitor research advancements in multimodal foundation models, implementing performance-driven optimizations with state-of-the-art techniques.
• Lead metric-focused industry-academia collaborations targeting computational efficiency and measurable outcomes (e.g., benchmark leadership).

Requirements

• MSc/PhD in Computer Science or equivalent research experience, with demonstrated achievements in multimodal pre-training architectures.
• Consistent research contributions evidenced by publications in top-tier conferences (e.g., ACL/ICML/NeurIPS) or equivalent industrial innovations.
• Expertise in training VLA/VLM/LLM systems using industry-standard frameworks (PyTorch/TensorFlow) with BERT/Transformer/GPT implementations.
• Proven experience in large-scale distributed training systems and production optimization techniques (e.g., model quantization).
• Familiarity with global research trends, with prior experience in globally leading AI research institutions considered advantageous.