音频多模态大模型评测与数据专家 - Data语音
Core
Design training data strategies and evaluation frameworks for multimodal audio large models to improve content creation and voice interaction capabilities.
Role type
Senior IC multimodal model evaluation and data strategy expert
Builds
Multimodal audio large models (speech synthesis, understanding, music generation) serving internal products (Doubao, TikTok, CapCut) and external enterprise clients via Volcano Engine.
Domain
AIGC, Multimodal Audio, Large Language Models
Deliverable
production ML models
Required skills
Multimodal model evaluation, training data strategy design, data quality assurance, evaluation metric definition, model capability decomposition, data annotation standarding, root cause analysis for model issues, cross-functional team leadership
Preferred skills
Linguistics, computational linguistics, Chinese linguistics, psychology, experience building evaluation systems from scratch, complex data production project leadership
Technologies
Large-scale audio models, multimodal AIGC frameworks, data annotation platforms, evaluation benchmarking tools
Responsibilities
Define model capability goals and training data strategies with algorithm and product teams; design and execute training data pipelines including analysis, strategy, standards, and quality acceptance; build and optimize objective and subjective evaluation systems for audio models; analyze evaluation results to identify model gaps and drive algorithm/data iterations; track industry advancements in multimodal data and evaluation methodologies; lead internal and external teams to execute data production and evaluation plans.
