电商场景LLM/VLM/AIGC推理工程师-TikTok Shop
Core
Design and implement a general-purpose LLM/VLM/AIGC inference engine for e-commerce scenarios, optimizing performance for real-time and batch processing across global GPU clusters.
Role type
Senior IC LLM/VLM inference engine developer
Builds
High-performance inference engines for e-commerce applications
Domain
E-commerce, Large Language Models, GPU Computing
Deliverable
production ML models
Required skills
C/C++, Python, Linux, Machine Learning frameworks (TensorFlow/PyTorch/MxNet), LLM/VLM/SD/DiT model optimization, Inference engines (vLLM/SGLang/TRT-LLM), GPU programming, Compiler optimization, Distributed storage, Cluster scheduling
Preferred skills
GPU hardware architecture understanding, CUDA/cuDNN stack expertise, GPU performance analysis, Hardware-software co-design, Deep framework internals knowledge
Responsibilities
Design and implement inference engines for online real-time and near-offline batch scenarios; Optimize inference performance for high concurrency, reliability, and scalability; Manage elastic scheduling and GPU overcommitment on global GPU clusters; Collaborate with algorithm teams for joint optimization
Seniority
Senior, hands-on IC