模型工程技术专家（AI Infra） - 剪映CapCut
Core
Deploying large-scale AI models (MoE, multimodal) to production and optimizing inference/training pipelines for video creation and GenAI products.
Role type
Senior Machine Learning Systems Engineer (AI Infrastructure)
Builds
Production-ready large language and multimodal models for CapCut, Dreamina, and related AIGC products.
Domain
Generative AI, Large Language Models, Video Creation
Deliverable
production ML models
Required skills
Python, PyTorch, C++, Linux, CUDA programming, GPU cluster optimization, vLLM, SGLang, model distillation, dynamic batching, quantization, attention mechanism optimization, distributed training (DeepSpeed, Megatron-LM), LoRA/QLoRA
Preferred skills
Agent-based workflows, automated evaluation, RLHF/SFT engineering, trillion-parameter model deployment
Responsibilities
Deploy large models to production using advanced inference frameworks; optimize inference performance via quantization and batching; build and optimize RL/SFT/End-to-End training pipelines; apply model distillation and data synthesis to reduce costs; debug GPU cluster bottlenecks and optimize operators.
Seniority
Senior, hands-on IC