1. Deeply engage in the R&D of multimodal large model foundations for embodied intelligence, optimizing model accuracy and training efficiency; 2. Deeply engage in developing visual capabilities for multimodal large models, including visual-video-language alignment, visual token compression, and enhancing spatio-temporal understanding; 3. Deeply engage in fine-tuning multimodal large models for embodied scenarios, leveraging techniques such as SFT, instruction fine-tuning, DPO, and GRPO for optimization; 4. Build high-quality training and testing datasets for embodied intelligence scenarios, establishing efficient evaluation metrics and data cleaning/annotation tools. 1、深度參與面向具身智能的多模態大模型基座的研發工作，優化大模型的精度與訓練效率； 2、深度參與多模態大模型的視覺能力研發，包括視覺視頻語言對齊，視覺Token壓縮，時空間理解能力提升等工作； 3、深度參與多模態大模型在具身場景的微調，利用包括SFT, 指令微調，DPO, GRPO等技術進行優化； 4、面向具身智能場景，構建高質量的訓練與測試數據集，搭建高效的評測標準與清洗標注工具。

1. Master the theoretical foundations and algorithms of artificial intelligence, pattern recognition, and deep learning; 2. Understand large-scale model technologies; preference given to candidates with publications in top-tier journals/conferences such as CVPR, ICCV, ICLR, ICML, and NeurIPS; 3. Proficient in C/C++ and Python programming, with familiarity in Linux and ROS systems; 4. Possess strong teamwork and communication skills, along with excellent learning capabilities. 1.掌握人工智慧、模式識別、深度學習理論基礎和演算法； 2.瞭解大模型相關技術，在CVPR，ICCV，ICLR，ICML，NeruIPS等TOP期刊會議發表論文者優先； 3.熟悉C/C++和Python編程，熟悉Linux系統和ROS系統； 4.具備良好的團隊協作能力和溝通能力，具備良好的學習能力。