腾讯混元-混元多模态数据产品（北京/上海）
Core
Building and managing datasets for pre-training and post-training of image/video multimodal large models.
Role type
Senior Multimodal Data Product Manager
Builds
Pre-training and post-training datasets for image/video generation models
Domain
AI / Computer Vision / Multimodal Large Models
Deliverable
production ML models
Required skills
Data strategy, data sourcing, quality filtering, data governance, cross-functional collaboration, SQL, Python, experimental design
Preferred skills
Experience with large-scale training data construction, pre-training data governance, post-training data production, image/video generation, multimodal LLMs, computer vision
Technologies
SQL, Python
Responsibilities
Define data standards and production plans for image/video model pre- and post-training; Source and evaluate public, licensed, collected, and synthetic datasets; Establish quality metrics for clarity, aesthetics, and consistency; Optimize data strategies based on training evaluation and error analysis; Collaborate with R&D to platformize data production and quality monitoring.