世界模型产品经理（模型训练方向） - AI数据与安全
Core
Product planning and strategy for world model training pipelines, focusing on multimodal data generation, simulation integration, and evaluation metrics to support AI foundation models and native applications.
Role type
Senior Product Manager (World Model Training & Data Strategy)
Builds
End-to-end world model training pipelines, multimodal datasets (text, image, video, 3D), and synthetic data generation strategies.
Domain
Artificial Intelligence, World Models, Multimodal Large Language Models, Computer Graphics
Deliverable
production ML models
Required skills
World model training pipeline design, multimodal dataset construction, SFT and RL strategy design, video generation model architecture understanding, game engine integration (UE/Unity), physical simulation alignment, evaluation metric development, cross-functional collaboration
Preferred skills
Video generation technology expertise (DiT, Sora), 3D asset management, physics engine simulation, academic research tracking
Technologies
DiT, Sora, Unreal Engine, Unity, Physics Engines
Responsibilities
Define end-to-end requirements for data collection, processing, training, and evaluation; design data flywheel processes for cleaning and labeling; explore synthetic data generation using video models and game engines; collaborate with researchers to establish evaluation benchmarks; track and translate academic advancements into product strategies.
