光子 AI-游戏模型数据专家-研发数据
Core
Process multimodal model data for game R&D, focusing on OCR, layout understanding, table/formula recognition, and image/video parsing to drive model performance.
Role type
Senior IC multimodal data engineer (game R&D)
Builds
High-quality, sustainable data flywheels for multimodal models
Domain
Gaming + Multimodal AI
Deliverable
production ML models
Required skills
Python, PyTorch, computer vision, NLP, multimodal understanding, large-scale data processing, distributed training, model evaluation
Preferred skills
Master's degree in CS/NLP/ML, algorithm R&D experience, multimodal LLM training flow knowledge (pre-training, SFT, RLHF/DPO)
Responsibilities
Process multimodal data for OCR, layout, tables, formulas, and media; lead data optimization via automated mining and cross-voting; design solutions for complex parsing scenarios; validate technical selections for frontier multimodal models.