机器学习研发工程师 - TikTok Search
Core
Design and implement machine learning and deep learning architectures for TikTok's high-concurrency search business, optimizing performance and cost for text relevance, multi-modal, and LLM models.
Role type
Senior IC machine learning engineer (search infrastructure)
Builds
High-performance search systems integrating generative search, SVLM ranking, and personalized recommendation
Domain
Internet search, large language models, deep learning engineering
Deliverable
production ML models
Required skills
Computer science fundamentals, data structures, algorithms, operating systems, deep learning frameworks, GPU programming, model training optimization, inference performance optimization, high-performance operator library design
Preferred skills
CUDA or Triton experience, vLLM/TRT-LLM/SGLang familiarity, FSDP/DeepSpeed/Megatron distributed training knowledge
Technologies
PyTorch, TensorFlow, CUDA, Triton, vLLM, TRT-LLM, SGLang, FSDP, DeepSpeed, Megatron
Responsibilities
Design and implement ML/DL architectures for search scenarios; Optimize performance and cost of deep learning models in high-throughput environments; Improve model training efficiency and accelerate inference performance; Design and develop high-performance operator libraries for GPU deployment