Responsibilities

1. Responsible for pre-training development and tuning of large language models and multimodal large models;

2. Responsible for robot control algorithm design, training, dataset processing, and real-machine deployment testing based on large models;

3. Responsible for improving large model training efficiency, such as model parallelism, flash attention, LoRA;

4. Responsible for tracking the latest field developments, pre-researching and evaluating large model applications in robotics, such as the RT series.

Requirements

1. Proficient in network structures and training methods related to large language models and multimodal large models;

2. Proficient in PyTorch/TensorFlow deep learning frameworks, master at least one of Python/C++;

3. Proficient in common distributed training frameworks and large model efficient fine-tuning techniques;

4. Proficient in using cloud computing for large-scale data processing and training;

5. Proficient in common imitation learning algorithms (e.g., ACT, DP) and Model-Based RL algorithms preferred;

6. Proficient in model and robot hardware interface deployment preferred;

7. Published papers in top AI, CV, NLP conferences/journals (e.g., TIP, TRO, CVPR, ACL) preferred.