腾讯游戏-大模型推理性能优化工程师/专家
Core
Design and optimize high-performance inference engines for Large Language Models (LLM), Vision-Language Models (VLM), and Diffusion Transformers (DiT) to enable efficient deployment.
Role type
Senior IC machine-learning inference optimization engineer
Builds
High-performance distributed inference systems for LLM/VLM/DiT models
Domain
AI/ML inference optimization on GPU and heterogeneous AI chips
Deliverable
production ML models
Required skills
C/C++, Python, distributed inference strategies (data/pipe-parallelism), GPU architecture, deep learning operator optimization, model architecture analysis
Preferred skills
NVLINK/GPU RDMA communication, low-bit quantization, PD separation, heterogeneous AI chip tuning
Technologies
vllm, sglang, tensorrt-llm
Responsibilities
Collaborate with algorithm teams to build industry-leading inference engines; Optimize inference performance via parallelism, low-bit computation, and operator tuning; Support mainstream GPUs and AI chips to maximize performance-cost efficiency.
