Software Engineer - Model Infrastructure, TikTok Feeds
Core
Design and develop scalable infrastructure for online inference and offline training of large-scale recommendation models to support complex foundation models.
Role type
Senior IC software engineer (model infrastructure)
Builds
Scalable compute and storage infrastructure for recommendation systems
Domain
Online advertising / Recommendation systems / GPU computing
Deliverable
infrastructure
Required skills
GPU architecture, CUDA, Triton Lang, PyTorch, TensorFlow, distributed model training, DNN compilers (MLIR, XLA, TVM)
Preferred skills
Low-level operator implementation, mathematical optimization of training/inference algorithms
Technologies
CUDA, CUTLASS, Triton Lang, TensorFlow, PyTorch, MLIR, XLA, TVM
Responsibilities
Build and optimize infrastructure for online inference and offline training; Solve bottlenecks in computation and storage; Co-design infrastructure with algorithm researchers; Improve model efficiency and scalability; Abstract reusable infrastructure components
Seniority
Senior, hands-on IC