多模态大模型应用算法专家 - Seed Model
Core
Building end-to-end multimodal data pipelines, reward systems, and evaluation frameworks to drive the performance of large multimodal models (MLLMs) for products like Doubao.
Role type
Senior IC multimodal algorithm engineer (data & RL)
Builds
Production multimodal models and agents for consumer apps and enterprise clients
Domain
AI / Machine Learning / Multimodal
Deliverable
production ML models
Required skills
Python, PyTorch, large-scale data cleaning, synthetic data generation, reward model design, RLHF/RLVR, multimodal evaluation, model training & tuning
Preferred skills
English reading/writing, mathematical foundations, data pipeline architecture, user feedback modeling
Technologies
Python, PyTorch
Responsibilities
Mining and synthesizing high-quality multimodal data; designing and optimizing reward models and verifiers for RL training; modeling user feedback signals from app usage; building multimodal evaluation systems; implementing multimodal understanding and agent capabilities.