Description
The role is a Video Generation Training Engineer responsible for building and operating a distributed training platform for a video generation foundation model. Responsibilities include creating reusable training infrastructure, managing job scheduling and checkpointed resumption, scaling training across hundreds of GPUs, building PB-scale video data pipelines, optimizing GPU memory and performance, and improving training stability. The position requires PyTorch distributed and CUDA expertise, familiarity with a mainstream training framework, hands-on experience with ≥256 GPUs, and experience building a training platform or toolchain.
